JP2014217091A - Information processing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the more highly accurate prediction scenario of a power demand even to a user who takes an uncertain action, and to optimize a power control plan, and to perform power control in accordance with the optimized plan.SOLUTION: An information processing system includes: first generation means for generating a plurality of prediction scenarios of a power demand in a target period as the target of prediction about each facility in a housing having facilities including a load device and a control target device; first calculation means for calculating the occurrence probability of each of the plurality of prediction scenarios; selection means for selecting the prediction scenarios one by one about each facility on the basis of the occurrence probability calculated by the first calculation means; and second generation means for generating the control plan of the facility on the basis of the combination of the prediction scenarios selected by the selection means.

Description

  The present invention relates to power control in a house.

  There is known a technique related to electric power control of a house called so-called HEMS (Home Energy Management System) (for example, Patent Documents 1 and 2 and Non-Patent Document 1).

JP 2012-100197 A JP 2007-295680 A

Shunsuke Ozoe, two others, “Optimization of Smart House Operation by Establishing Mixed Integer Programming with Recourse”, IEEJ Transactions B, Vol. 131, No. 11, pp. 885-895, November 2011

One of the objects of the present invention is to provide a power supply / demand prediction scenario with higher accuracy even for a user who performs an uncertain action.
Another object of the present invention is to provide a power supply / demand prediction scenario that reflects a predetermined user's behavior.
Yet another object of the present invention is to provide a HEMS having the new value of reducing carbon dioxide emissions.
Yet another object of the present invention is to provide more flexible power control according to user demands.

  A first aspect of the present invention is in a house having a facility including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. A first calculation means for calculating a plurality of prediction scenarios of power supply and demand and the occurrence probability of each of the plurality of prediction scenarios for each of the facilities using statistical processing of past actual values in each of the facilities; The second calculation means for calculating the probability that a combination of specific prediction scenarios will occur at the same time in the facility using the occurrence probability calculated by the first calculation means, and the occurrence probability calculated by the second calculation means Provides an information processing system having an output means for outputting a combination of prediction scenarios having the largest value as an integrated scenario.

  A second aspect of the present invention is in a house having equipment including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. An acquisition means for acquiring a use schedule for at least one of the facilities, and a generation means for generating a power supply / demand prediction scenario for each of the facilities, reflecting the use schedule acquired by the acquisition means. An information processing system is provided.

  A third aspect of the present invention relates to an electric device, a power supply device, a heat supply device, and a house having a facility including one or more devices serving as power loads and one or more devices to be controlled among electric vehicles. Generating means for generating a power supply / demand prediction scenario for each of the facilities, and using the prediction scenario generated by the generating means, the electric device, the power supply device, the heat supply device, and the electric vehicle. 1 or more optimization means for optimizing the control plan, wherein the optimization algorithm includes a first mode for minimizing carbon dioxide emission for generating electric power used in the facility. An optimization unit including a plurality of algorithms corresponding to the operation mode and performing the optimization according to one operation mode selected from the one or more operation modes; An information processing system comprising: generation means for generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on a control plan optimized by I will provide a.

  A fourth aspect of the present invention is in a house having a facility including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. Generating means for generating a power supply / demand prediction scenario for each of the facilities, and using the prediction scenario generated by the generating means, the electric device, the power supply device, the heat supply device, and the electric vehicle. Optimization means for optimizing the control plan by a probability programming method using an objective function, wherein the optimization algorithm includes a plurality of algorithms corresponding to a plurality of operation modes each using a different objective function , An optimization means for performing the optimization according to one operation mode selected from the plurality of operation modes, and a control plan optimized by the optimization means Based on the electrical device, the power supply device, an information processing system having a generating means for generating control rules for controlling at least one of the heat supply device and the electric vehicle.

  A fifth aspect of the present invention is a first generation for generating a plurality of prediction scenarios of power supply and demand in a target period to be predicted for each of the facilities in a house having facilities including a load device and a control target device. Means for each of the plurality of prediction scenarios, one for each of the facilities based on the occurrence probability calculated by the first calculation means, and a first calculation means for calculating the occurrence probability of the prediction scenario. There is provided an information processing system comprising selection means for selecting a prediction scenario and second generation means for generating a control plan for the facility based on a combination of prediction scenarios selected by the selection means.

According to the first aspect, it is possible to provide a power supply / demand prediction scenario with higher accuracy even for a user who performs an uncertain action.
According to the second aspect, it is possible to provide a power supply / demand prediction scenario that reflects a predetermined user action.
According to the third aspect, it is possible to provide a HEMS having a new value of reducing carbon dioxide emissions.
According to the 4th aspect, more flexible electric power control according to a user's request can be provided.
According to the fifth aspect, it is possible to provide power control based on the supply and demand prediction with higher accuracy.

It is a figure which illustrates the object of prediction and control by a control algorithm. It is a figure which illustrates the subject of charge and discharge control of EV and a storage battery. It is a figure which illustrates the composition of HEMS concerning one embodiment. It is a figure which illustrates the apparatus structure in the smart house which concerns on one Embodiment. It is a figure which illustrates the effect by control of each apparatus. It is a figure which illustrates optimization mode. It is a figure which illustrates plan time. It is a figure which illustrates plan time. It is a figure which illustrates the operation | movement flow of HEMS. An input / output example of the global optimum plan is shown. An input / output example of the global optimum plan is shown. An input / output example of the global optimum plan is shown. An input / output example of the global optimum plan is shown. An input / output example of the global optimum plan is shown. An input / output example of the global optimum plan is shown. It is a figure which illustrates the daily process of a whole optimal plan. It is a figure which illustrates the recalculation process of the whole optimal plan at the time of a connection / release. It is a figure which illustrates the process for every hour of a control rule set. It is a figure which illustrates the process for every minute of an execution instruction. It is a figure which illustrates the module list of a control algorithm. It is a figure which illustrates the module structure of a global optimal plan. It is a figure which illustrates the flow of processing of each module. It is a figure which illustrates the submodule list of a demand module. The submodule structure of a demand module is shown. It is a figure which illustrates the flow of the submodule of a demand module. The submodule list of a supply module is shown. It is a figure which illustrates the submodule structure of a supply module. It is a figure which illustrates the flow of the submodule of a supply module. The submodule list of an electrical storage module is shown. It is a figure which illustrates the module structure of an electrical storage module. It is a figure which illustrates the flow of the submodule of an electrical storage module. It is a figure which illustrates the submodule structure of the optimal supply-and-demand plan module. The submodule configuration of the optimal supply and demand planning module is shown. It is a figure which illustrates the flow of the submodule of an optimal demand-and-supply planning module. It is a figure which illustrates the category of a demand basic pattern. It is a figure explaining the household appliance power demand scenario creation flow. It is a figure which illustrates the relationship with the temperature of the household electrical power demand for every season. It is a figure which illustrates the list of the variable and the symbol used by the input / output of a household appliance power demand scenario creation submodule submodule, and a model. It is a figure which illustrates the variable and the symbol list of a household appliance power demand scenario creation submodule. It is a figure which illustrates the relationship with the temperature of the household electrical power demand in the summer and winter. It is a figure which illustrates the relationship with the temperature of the household electrical power demand in the spring and autumn. It is a figure which shows another example of the relationship with the temperature of the household electrical power demand in the spring and autumn. It is a figure which illustrates the representative value obtained from error distribution. It is a figure explaining the EV electric power demand scenario creation flow. It is a figure which illustrates EV electric power demand scenario creation submodule input / output. An example in which four swing widths are set is shown. It is a figure explaining a heat demand scenario creation flow. It is a figure which illustrates the input / output of a heat demand scenario creation submodule. It is a figure which illustrates the representative value obtained from error distribution. It is a figure explaining PV supply amount scenario creation flow. It is a figure which illustrates the input / output of PV supply amount scenario creation module. It is a figure which illustrates the variable and symbol list of PV supply amount scenario creation module. It is a figure which illustrates the relationship between solar radiation amount and power generation efficiency. It is a figure which illustrates the relationship between PV power generation efficiency and temperature. It is a figure which illustrates the amount of solar radiation by the weather. It is a figure which illustrates the representative value obtained from error distribution. It is a figure which illustrates the input / output in a system power supply submodule. It is a figure which illustrates the input-output in a fuel cell submodule. It is a figure which illustrates the input-output in a thermal storage unit calorie | heat amount submodule. It is a figure which illustrates the input-output in EV electric power storage amount submodule. It is a figure which illustrates the input-output in a storage battery electrical storage amount submodule. It is a figure explaining the flow of an integrated scenario creation submodule. It is a figure which illustrates the input / output of an integrated scenario creation submodule. It is a figure which illustrates the list of the variable and the symbol of an integrated scenario creation submodule. It is a figure explaining the production | generation flow of the co-occurrence probability of the scenario which assumed the multidimensional normal density function. It is an image figure when there are five scenarios of the PV supply amount prediction model, the home appliance power demand prediction model, and the heat demand prediction model. It is a figure which illustrates the production | generation of an integrated scenario. It is a figure which illustrates the item of an integration scenario. It is a figure which illustrates a parameter item. It is a figure which illustrates a parameter item. It is a figure which illustrates a parameter item. It is a figure which illustrates a parameter item. It is a figure which illustrates the output of this submodule. It is a figure which illustrates the list of the variables used by the optimization model. It is a figure which illustrates the list of the variables used by the optimization model. It is a figure which illustrates the list of the variables used by the optimization model. It is a figure which illustrates the structure of the apparatus considered by optimization. It is a figure which illustrates the linear formula which comprises an objective function. It is a figure which illustrates the setting method of the input parameter for every optimization mode. It is a figure which illustrates a response | compatibility with the output item and the variable used by optimization. It is a figure which illustrates the input / output of a control rule submodule. It is a figure which illustrates the basic state of a device. The basic state is an example of control during discharge. The basic state is an example of control during charging. It is a figure which illustrates EV storage battery storage plan value and a basic state. It is a figure which illustrates the control rule of the storage battery created from FIG. It is a figure which illustrates control of the fuel cell for every optimization mode. It is a figure which illustrates control of a fuel cell in case a storage battery is a discharge state. It is a figure which illustrates control of a fuel cell in case a storage battery is a charging state. It is a figure which illustrates cooperation of this system and its peripheral system. It is a figure which illustrates the conceptual function division structure of a control algorithm. It is a figure which illustrates the functional division of a control algorithm. It is a figure which illustrates the functional division subdivided. It is a figure which illustrates the physical function of a system, its function, and a role. It is a figure which illustrates a system block diagram. It is a figure which illustrates the sequence diagram which concerns on a control rule production | generation process. It is a figure which illustrates the event which may become a processing opportunity. It is a figure which illustrates the argument of control rule generation processing starting. It is a figure which illustrates an acceptance notification state. It is a figure which illustrates an acceptance notification state. It is a figure which illustrates an acceptance notification state. It is a figure which illustrates the data flow in this system. It is a figure which illustrates the main process IO related figure. It is a figure which illustrates a processing order flow chart. It is a figure which illustrates an error check list. It is a figure which illustrates an acquisition data list. It is a figure which illustrates a stored data list. It is a figure which illustrates an update data list. It is a figure which illustrates the flow of the main process. It is a figure which illustrates processing status update contents. It is a figure which illustrates processing status update contents. It is a figure which illustrates the reference source and storage destination of input data. It is a figure which illustrates the reference source and storage destination of input data. It is a figure which illustrates the reference source and storage destination of input data. It is a figure which illustrates the reference source and storage destination of input data. It is a figure which illustrates a processing order flow figure (repost). It is a figure which illustrates a processing status classification list. It is a figure which illustrates processing status update contents. It is a figure which illustrates household appliance power demand scenario creation IO related figure. It is a figure which illustrates the argument of a household appliance power demand scenario creation function. It is a figure which illustrates the return value of a household appliance power demand scenario creation function. It is a figure which illustrates error check specification. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates storing of the value to a database. It is a figure explaining the household appliance power demand scenario creation processing flow. It is a figure which illustrates the relationship between household appliance electric power demand and temperature in spring and autumn. The figure which shows another example of the relationship between household appliance electric power demand and temperature in spring and autumn. It is a figure which illustrates IO related figure of EV electric power demand scenario creation function. It is a figure which illustrates the argument of EV electric power demand scenario creation function. It is a figure which illustrates the return value of EV electric power demand scenario creation function. The figure which illustrates the error check specification of EV electric power demand scenario creation function. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates storing of the value to a database. It is a figure explaining the EV electric power demand scenario creation process flow. It is a figure which illustrates a heat demand scenario creation function IO related figure. It is a figure which illustrates the argument of the heat demand scenario creation function. It is a figure which illustrates the return value of a heat demand scenario creation function. It is a figure which illustrates the error check specification of a heat demand scenario creation function. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates storing of the value to a database. It is a figure which illustrates a heat demand scenario creation processing flow. It is a figure which illustrates PV supply scenario creation function IO related figure. It is a figure which illustrates the argument of PV supply scenario creation function. It is a figure which illustrates the return value of a PV supply scenario creation function. It is a figure which illustrates the error check specification of PV supply scenario creation function. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates storing of the value to a database. It is a figure explaining a PV supply scenario production | generation process flow. It is a figure which illustrates IO related figure of an integrated scenario creation function. It is a figure which illustrates the argument of an integrated scenario creation function. It is a figure which illustrates the return value of an integrated scenario creation function. The error check specifications of the integrated scenario creation function are shown below. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates storing of the value to a database. It is a figure explaining an integrated scenario creation processing flow. An example of processing is shown. An implementation example is shown. It is a figure which illustrates the IO related figure of optimization plan preparation. It is a figure which illustrates the argument of the optimization plan creation function. It is a figure which illustrates the return value of an optimization plan creation function. It is a figure which illustrates the error check specification of an optimization plan creation function. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates storing of the value to a database. It is a figure which illustrates storing of the value to a database. It is a figure which illustrates storing of the value to a database. It is a figure which illustrates a processing list. It is a figure which illustrates the list of the variables required in order to define an optimization model. It is a figure which illustrates the list of the variables required in order to define an optimization model. It is a figure which illustrates the list of the variables required in order to define an optimization model. An implementation example is shown. It is a figure which illustrates a constraint condition. It is a figure which illustrates a constraint condition. It is a figure which illustrates a constraint condition. Implementation example. It is a figure which illustrates the linear formula which comprises an objective function. Implementation example. It is a figure which illustrates the risk parameter and variable corresponding to an output item. It is a figure which illustrates the risk parameter and variable corresponding to an output item. An example of obtaining a variable x [t] is shown. It is a figure which illustrates IO figure of a control rule creation function. It is a figure which illustrates the argument of the optimization plan creation function. It is a figure which illustrates the return value of an optimization plan creation function. It is a figure which illustrates the error check specification of an optimization plan creation function. It is a figure which illustrates acquisition of a value from a database. It is a figure which illustrates storing of the value to a database. It is a figure explaining a control rule production | generation processing flow. It is a figure which illustrates control to a device. It is a figure which illustrates the basic state (at the time of system power supply connection) of a device. It is a figure which illustrates the basic state (at the time of disconnection from a system power supply) of a device. The example of EV storage battery storage plan value and a basic state is shown. It is a figure which illustrates the control rule of the storage battery created from FIG. It is a figure which illustrates optimization mode and control of a fuel cell. It is a figure which illustrates control of a fuel cell in case a storage battery is a discharge state. It is a figure which illustrates control of a fuel cell in case a storage battery is a charging state. The items copied to the time series optimization parameter table among the setting items of the house master are shown. The items copied to the time series optimization parameter table among the setting items of the house master are shown. The items copied to the time series optimization parameter table among the setting items of the house master are shown. Of the setting items of the house master, items that are copied to the optimization parameter table are shown. Of the setting items of the house telemaster, items to be copied to the optimization engine parameter table are shown. It is a figure which illustrates the precondition of system operation. The parameters that can be specified in this system are shown below. The parameters that can be specified in this system are shown below. The parameters that can be specified in this system are shown below. It is a figure which illustrates a table list. It is a figure which illustrates a table list. It is a figure which illustrates a table list. It is a figure which illustrates the calculation formula and approximate value of the number of records stored in each table. It is a figure which illustrates the calculation formula and approximate value of the number of records stored in each table. It is a figure which illustrates the value of the parameter used for calculation of an approximate value. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. It is a figure which illustrates the data item list of each table. 1 is a diagram illustrating a device configuration of a power control system 1 according to an embodiment. 1 is a diagram illustrating a functional configuration of a power control system 1. FIG.

1. 1. Introduction Considering energy supply and demand, it will become increasingly important to reduce carbon dioxide emissions and promote energy conservation. In order to realize this, it is considered that energy-generating devices such as solar power generation, electric vehicles, fuel cells, etc., and storage batteries for homes will be widely used. The spread of such energy creation devices and energy storage devices to homes provides consumers with options for procuring and using energy, and at the same time consumers are able to discharge electricity from purchased electricity, solar power generation, fuel cells, storage batteries and electric vehicles. If it is not managed properly, it will lead to a loss situation and the need to optimally control the creation of energy and the use of energy storage devices in the home. However, energy creation devices and energy storage devices have capacity limitations, power generation and heat generation characteristics, and it is impossible to operate them optimally according to demands that vary depending on the climate and seasons. In addition, the management of energy supply and demand in houses is a task that did not exist in the conventional living environment so far, and it is possible to create and store energy without bothering the people living in such houses. Implementation of a control algorithm that optimally controls energy equipment is essential. Based on this background, this control algorithm is suitable for energy generation equipment and energy storage equipment in a house with grid power supply, photovoltaic power generation, electric vehicle or residential storage battery, and fuel cell. The purpose is to control.

  In this embodiment, the control plan of the equipment in the house is optimized by the probability programming method.

2. Definition of terms (1) Power supply System power supply, PV, EV, storage battery, or fuel cell.
(2) System power supply Power supplied from the commercial distribution network owned by the power company.
(3) Energy creation equipment Equipment that creates usable power and effectively uses energy. For example, solar cell panels and fuel cells.
(4) PV
Photovoltaic panel.
(5) Energy storage equipment A system that stores power and supplies it when needed. For example, storage batteries and EV built-in batteries.
(6) EV
Electric vehicle. For example, one that can be charged and discharged in both directions including Vehicle to Home (V2H).
(7) Storage battery A stationary storage battery for home use.
(8) Fuel cell A battery that extracts electric power through an electrochemical reaction. It may be abbreviated as FC.
(9) HEMS
Home Energy Management System.
(10) ECHONET Lite or Echonet Lite
Abbreviation for Energy Conservation and Homecare Network. A standard communication protocol that realizes a highly reliable home network with a simple telegram structure.
(11) Control algorithm In a detached smart house, an algorithm that optimally controls the balance between power supply by a distributed power source of a full set of system power supply, PV, EV, and fuel cell and the power demand of household devices. Includes global optimal planning, control rule set, and plan correction. “Algorithm” and “optimization algorithm” are also synonymous.
(12) Definition of control algorithm requirements Definition of control algorithm outline, model structure (diagrams and mathematical expressions), processing flow, variable list (input and output), and term list.
(13) Control algorithm design Based on the control algorithm requirement definition, design control algorithm outline, configuration, development items, functions, methods, test policies, external interface, initial processing, etc. It is divided into a general design related to input / output of each submodule and a basic design related to the interior of the submodule.
(14) Variable A numerical value or character string that changes with time. For example, household electrical power demand.
(15) Parameter A numerical value or character string set by the system administrator or user. It may be different for each house, and may be updated manually at regular or irregular intervals. For example, electricity charges.
(16) Optimization mode A combination of objective function, constraint conditions, and parameters used to optimize power supply and demand. One that can be selected from a plurality of combinations. For individual optimization modes, see 3.4. See optimization mode.
(17) Optimization problem The problem of maximizing or minimizing a real or integer function.
(18) Objective function A function that is minimized or maximized in an optimization problem.
(19) Optimization calculation Calculation processing for solving an optimization problem.
(20) Restriction An equality or inequality that the solution must satisfy in an optimization problem.
(21) Optimization solver Numerical computation software that performs optimization calculations with the constraints of the optimization problem and the objective function as inputs.
(22) HEMS subsystem An information management server, smartphone, home server, device network, and sensor network that physically configure the HEMS.
(23) Information management server HEMS cloud platform.
(24) Smartphone A mobile communication terminal installed in a detached smart house. The main user interface.
(25) Home server A hardware interface for connecting an information management server, a device network, and a smartphone. The rule condition received from the information management server is determined, and an execution instruction is transmitted to the device network.
(26) Device Refers to devices that supply, supply, and store electricity.
(27) Equipment network
A communication network consisting of power supplies and home appliances that run on ECHONET Lite.
(28) Sensor network A communication network consisting of sensors such as operating states of various devices and temperature and humidity meters. Positioning as a network for collecting measurement data, EV external data centers, weather information servers, and EV reservation inputs may be handled at the same level.
(29) Optimization engine I / F
A function that mediates request calls between the optimization engine and other modules.
(30) Configuration unit Refers to the overall optimum plan, control rule set, execution instruction, execution, and monitoring that logically configure the HEMS.
(31) Overall optimum plan A component that is installed on the information management server and includes a control algorithm. Based on the result data, prediction, optimization calculation, and control rule generation are performed.
(32) DB for analysis
Although it is not a module for the global optimum plan, it is a component installed on the information management server, and exchanges data with the global optimum plan. Actual data storage, prediction data storage, optimization plan data storage, control rule storage, parameter registration, individual control rule registration, control rule subset extraction, total optimal plan call, home server and total optimal plan Exchange data with.
(33) Execution instruction A control algorithm component installed on the home server. Judge the conditional expression of the control rule subset, convert the execution instruction that satisfies the conditional expression to ECHONET Lite, and send it to the device network.
(34) Execution Unless otherwise specified, indicates that various devices operate.
(35) Monitoring Measurement of performance data such as operating status and temperature / humidity of various devices. With the positioning of collecting measurement data, EV power demand results and weather information collected by external servers, and EV reservations input to smartphones may be handled at the same level.
(36) Control rule A rule that specifies the operation of various devices with respect to time or conditional expressions. For example, it is described by an IF THEN statement.
(37) Control rule subset A subset extracted from the control rule set by priority and activation time.
(38) EV data center A data center of EV performance values operated by an automobile manufacturer.
(39) Weather information service A service that provides weather information. For example, provided by an external operator.
(40) Function category This refers to the main function category, demand function category, supply function category, power storage function category, and optimum supply and demand plan function category that constitute the overall optimum plan.
(41) Main function category Function category that realizes integration of other function categories and cooperation with external systems.
(42) Demand function category A function category that calculates a predicted value based on the actual value and calculates a prediction error from the actual value and the predicted value for household electrical power demand, EV power demand and hot water demand in the overall optimum plan.
(43) Supply function category A function category for calculating a predicted value based on the actual value and calculating a prediction error from the actual value and the predicted value for the PV supply amount and the heat storage unit heat amount in the overall optimum plan.
(44) Power storage function category A function category for calculating a prediction value based on the actual value and calculating a prediction error from the actual value and the prediction value for the EV power storage amount and the storage battery power storage amount in the overall optimum plan.
(45) Optimal supply and demand planning function category In the overall optimization plan, based on the calculation results and parameters of other function categories, the planned value of the operation status of the equipment that maximizes or minimizes the objective function set in the optimization mode A functional category that calculates control rules for generating further planned values.
(46) Module Refers to a demand module, a supply module, a power storage module, and an optimum supply and demand planning module that constitute an overall optimum plan.
(47) Demand module A module that calculates a predicted value based on the actual value and calculates a prediction error from the actual value and the predicted value for household electrical power demand, EV power demand, and heat demand in the overall optimum plan. Calculation result variables can be called from other globally optimal planning modules.
(48) Supply module A module that calculates a prediction value based on the actual value for the PV supply amount and the heat storage unit heat amount in the overall optimum plan, and further calculates a prediction error from the actual value and the predicted value. Calculation result variables can be called from other globally optimal planning modules.
(49) Power storage module A module that calculates a predicted value based on the actual value and calculates a prediction error from the actual value and the predicted value for the EV charged amount and the stored battery charged amount in the overall optimum plan. Calculation result variables can be called from other globally optimal planning modules.
(50) Optimal supply and demand planning module Based on the calculation results and parameters of the other modules in the overall optimization plan, the planned operating conditions of the equipment that maximize or minimize the objective function set in the optimization mode are calculated. And a module for generating a control rule for realizing the planned value. Calculation result variables can be called from other globally optimal planning modules. The generated control rule is transmitted to the control rule set of the analysis DB. It also holds the main calculation results of the global optimum plan and can call balance formulas and variables from other modules and external systems.
(51) Mesh The step size (unit) in the overall optimal plan.
(52) Span The length of time in the overall optimum plan.
(53) Cycle The update frequency of the overall optimum plan.
(54) Heat storage unit A hot water storage tank constituting the fuel cell. The amount of heat stored in the heat storage unit is called the heat storage unit heat amount.
(55) Basic demand pattern A demand curve for 24 hours according to the day of the week. Several types, such as morning and evening types and night types, are prepared as initial values, and after the actual values are obtained, they are replaced with the median of the actual values.
(56) Percentile value Percentile or percentage point. Boundary value when the whole is divided into 100 by frequency distribution. The 0th percentile value is the minimum value, the 50th percentile value is the median value, and the 100th percentile value is the maximum value. In addition, 5, 25, 75, or 95 percentile values are often used.
(57) Percentile point The actual numerical value located at the percentile value.
(58) Percentile section A section specified by two percentile values.
(59) Prediction error The difference obtained by subtracting the actual value from the predicted value. There is a different value for each predicted point in time for one actual value. Further, by accumulating the prediction error, it is possible to generate a prediction error distribution (not limited to a normal distribution) and a percentage point.
(60) Demand scenario A curve in which demand forecast values are connected in the future along the time axis. In theory, there can be an infinite number of scenarios for one actual value. In this control algorithm, a plurality of% points are extracted based on the distribution of prediction errors to generate a plurality of scenarios.
(61) Supply scenario A curve obtained by connecting the predicted supply values with possible values along the time axis. Same as demand scenario.
(62) Integrated scenario A combination of demand and supply scenarios.
(63) Occurrence probability The probability that the numerical value according to the demand scenario, supply scenario, and integration scenario will actually occur.
(64) Amount of safe power storage Amount of power stored in excess to cope with uncertainties in demand and to prevent battery exhaustion.
(65) Management width The upper and lower limits of the amount of stored electricity determined to be appropriate through optimization.
(66) Graph generator A module installed on the information management server. The actual value, the predicted value, the planned value, and the prediction error held by the overall optimal plan are displayed in a graph.
(67) Risk parameter A parameter for converting the risk of prediction error into cost. A large setting is a risk-avoidance plan, and a small setting is a risk-loving plan.
(68) Model A phenomenon or behavior expressed using mathematical expressions.
(69) Category For each house, past household electrical power demand is classified by day of the week, public holidays and weather.
(70) Recourse cost Penalty cost given to a power shortage or surplus that would occur when the amount of demand for home electric power and the amount of PV supply assumed in the scenario are realized.
(71) LDV (limited dependent variable) model Limited dependent variable model. It is a model that estimates in consideration of limited observation information and missing structure when the dependent variable is in an observable missing state. Also called the Tobit model. If the occurrence of a deficiency is not random, if the generation structure of the deficient data is ignored and the estimation is performed, a bias is generated in the estimator, and it is necessary to estimate with such a model.
(72) Uniform random numbers Random numbers that divide a certain finite interval and that all real numbers appear in the interval with the same probability (concentration).
(73) Objective variable A variable that represents the result of a thing. Variable to be predicted.
(74) Explanatory variable Variable that seems to be the cause of things. Variable used to calculate the predicted value.
(75) Linear regression model A model that estimates a coefficient and an intercept term of an independent variable on the assumption that the dependent variable can be expressed by a linear function expression with respect to the independent variable.
(76) Single regression A linear regression model that has one explanatory variable.
(77) Sensitivity An index indicating how much one variable reacts to the other for two variables.
(78) Likelihood An index that indicates how easily observed data is seen from the model assumed as a precondition.
(79) Log-likelihood The logarithm of likelihood.
(80) Normal distribution Gaussian distribution. Distribution of the likelihood of the value of a variable obtained when infinitely repeated binary independent trials with the same probability as in the front and back of a coin. Random variables expressed as the sum of many independent factors follow a normal distribution. A normal distribution curve obtained by graphing a normal distribution is a symmetrical curve with a symmetrical shape, and is also called a bell curve because it resembles a bell shape.
(81) Standard normal distribution Normal distribution with mean 0 and variance 1.
(82) User Mainly refers to a resident of a HEMS house, and sets a HEMS house according to usage.
(83) Administrator A person who maintains and operates an information management server including this system.

3. Outline 3.1. Challenges in energy utilization optimization 3.1.1. Basic concept of the control algorithm In the control algorithm, the basic concept of the control target is set as follows.
(1) Regarding power supply and demand in a house, it is desirable that a consumer can enjoy an economic merit (that energy can be used at a lower cost than the case without control).
(2) Prioritize the use of electric power generated by solar power generation and make maximum use of it for the demand for electric power in the house.
(3) The control of the energy creation device and the energy storage device in the house is performed by predicting the energy supply and demand of the house in the future in advance so as not to disturb the convenience of consumers.
(4) The future power supply and demand situation in the house predicted by the control algorithm increases in accuracy according to past data accumulation.

3.1.2. Problem of Energy Utilization Optimization FIG. 1 is a diagram illustrating an example of prediction and control targets by a control algorithm. In this control algorithm, necessary predictions are made as needed for (1) to (6) in FIG. 1 in accordance with the basic concept described above in terms of energy generation in the house, power generation, charging, and heat supply by energy storage devices. , (2) to (6) are controlled. The issues for realizing optimal control based on energy independence in the home by the control algorithm are described below for each home appliance power demand in the home and each control target device.

(1) Home appliance power and heat demand The pattern of home appliance power demand and heat demand in a house varies depending on whether the day is a weekday or a holiday (including holidays). For this reason, it is necessary to supply power that cannot generate electricity from PV (home appliance power demand and PV generated energy) from FC or EV discharge. It is necessary to consider the economic efficiency at a certain point. In addition, power generation from FC is closely related to heat supply, and the use of only generated power greatly reduces the energy utilization efficiency. For this reason, it is required to predict the transition of home electric appliance power demand and heat demand to the future of the day of the control, which is considered to affect the control of the day. In addition, the variables used for the prediction are required to be automatically adjusted to higher accuracy by being corrected as the result data is accumulated.

(2) PV
The amount of power generated by PV is not a control target. However, it is required to predict the difference with the demand for household electric appliances in the future. Since the amount of power generated by PV greatly depends on the weather (the amount of solar radiation), it is necessary to predict the amount of power generated in the future in consideration of the future weather and the relationship with the amount of power generated by PV due to weather. In addition, regarding the prediction of the amount of generated power of PV, it is required that the variable used for the prediction is corrected as the result data is accumulated and automatically adjusted to a higher accuracy.

(3) FC
FC is a device that achieves high energy use efficiency through cogeneration (combination of electric power and heat), and is a valuable source of electric power and heat for realizing self-sufficiency of residential energy. When controlling the amount of electric power generated from FC, it is desirable to meet the heat demand required for the house. For this reason, when controlling the amount of power generated from the FC, it is required to perform an appropriate operation so that sufficient heat can be secured at a necessary time so as not to cause a shortage of heat in the house. For this purpose, it is necessary to predict the heat demand of the house. Regarding the prediction of the amount of power generated by FC, it is required that the variables used for the prediction are corrected as the result data is accumulated and automatically adjusted to a higher accuracy.

(4) EV and (stationary) storage battery The EV storage battery is expected to be used as a storage battery to fill the difference between the demand for home electric power and the PV power generation as a storage battery. However, such EVs cannot be used when EVs are used as automobiles and leave homes. In addition, if the EV is excessively discharged, the amount of power necessary for traveling cannot be secured, which greatly hinders the convenience of consumers. Furthermore, it is necessary from the viewpoint of economy that the timing of discharging from the EV or the storage battery is performed mainly in the daytime period when the charge of purchased power is high. For this reason, it is calculated | required that the usage pattern of EV which differs for every house is estimated, and the charge / discharge amount which can be anticipated for EV is controlled appropriately. Regarding EVs, the variables used for predicting their use are required to be corrected as the performance data is accumulated and automatically adjusted to higher accuracy. In the case of a (stationary) storage battery, as in the case of EV, it is necessary to secure a minimum capacity necessary for power use at the time of a power failure and to perform appropriate charge / discharge control in consideration of the discharge timing.

  FIG. 2 is a diagram illustrating a problem of EV and storage battery charge / discharge control.

3.2. Overall Configuration FIG. 3 is a diagram illustrating a configuration of a HEMS according to an embodiment. In this example, the HEMS includes an information management server, a cradle server, a terminal device, a control target device (controlled device), and a sensor network. The terminal device is a PC, a tablet, or a smartphone. The device to be controlled is a battery, EV, or fuel cell. The sensor network is a power and gas sensor. The terminal device, the device to be controlled, and the sensor network are provided in the house and connected to the cradle server. In this example, operation control of load equipment such as an air conditioner is not performed.

  FIG. 4 is a diagram illustrating a device configuration in the smart house according to the embodiment.

3.3. FIG. 5 is a diagram exemplifying the effect of the control of each device. When each device is controlled based on the concept and problem of the control algorithm described above, it is expected that the effect shown in FIG. It is desirable that the control algorithm can be expected to achieve this level of effect.

3.4. Optimization mode One of the features of the smart house according to the present embodiment is self-sufficiency. However, optimization can have multiple objectives, such as self-sufficiency, economy and environment. In order to achieve such a plurality of purposes, the control algorithm in the present embodiment prepares an “optimization mode” for each purpose, and the system administrator or user can set the mode to be used. The optimization mode is a combination of an objective function, constraint conditions, and parameters used for optimizing power supply and demand. One combination can be selected from a plurality of combinations. Various optimal controls can be realized by setting cost and penalty parameters for each optimization mode.

  FIG. 6 is a diagram illustrating an optimization mode. Here, seven modes of economy (saving), self-sufficiency, environment (eco), charging priority, silent (good night), long battery life, and emergency (degeneration) are illustrated. Economic (saving) mode is a mode that minimizes power and fuel costs. Since PV has no cost, PV is given top priority. In addition, according to the price plan and time zone, the cheaper one of the system power supply and the fuel cell is compared, and the cheaper power is always used. Pursuing economic merits as described above.

  The self-sufficiency mode is one of the most important optimization modes. In order to realize independent power supply and demand for each house as much as possible without depending on the grid power supply, the amount of power purchased from the grid power supply is minimized. In addition to making the best use of PV, we aim to achieve independent power supply and demand even during normal times by using fuel cells for nighttime and peak power demand. In this embodiment, this self-sufficiency mode is set as the initial value of the optimization mode.

  The environment (eco) mode is a mode for minimizing carbon dioxide emissions. Prioritize PV with zero carbon dioxide emissions. In addition, we will give priority to the one with lower carbon dioxide emissions among the system power supply and fuel cell, aiming for environmentally friendly power supply and demand.

  The charge priority mode is a mode in which the remaining battery level is maximized so as to reduce the fear of running out of the battery as much as possible. For example, it can be set so that it can always depart at full charge in an EV installation home. It is also effective when there are concerns about power outages. However, it should be noted that in the state of full charge during the day, the electric power generated by PV cannot be stored and all flows backward, so that there is a large loss in terms of self-sufficiency and cost.

  The quiet mode is a mode in which the nighttime volume is suppressed by minimizing the switching operation of the fuel cell and the PCS at nighttime. Aim to sleep at home and consider the neighborhood.

  The battery long-lasting mode is a mode for preventing deterioration of the battery by setting a penalty for the number of charge / discharge switching times.

  The emergency (degenerate) mode is a mode that maximizes the self-sustaining operation period when the system power is disconnected. Combined with the three power sources of PV, fuel cell, and EV / storage battery under the constraint that the power flow / reverse power flow of the system power supply is zero but fuel can be supplied, the fuel cell is utilized without wasting as much power generation as possible. However, the time difference between supply and demand is absorbed by the EV / storage battery. Specifically, a penalty value such as battery exhaustion is specially set for the normal self-sufficiency mode. However, it should be noted that there is a limitation in cooperation between hardware, such as PV cannot generate power when there is no EV.

3.5. Planning Time FIGS. 7 and 8 are diagrams illustrating the planning time. There are three planned times: mesh, span, and cycle. These are often confused and will be described here. The mesh is a step size in the global optimum plan. In the control algorithm in this embodiment, the time interval is 1 hour. However, in the present embodiment, in consideration of future expandability, the data structure is held with a 10-minute mesh. This is because most of the current weather information is a mesh for a minimum of one hour, and if the mesh is too fine, it is assumed that the amount of data increases and the communication load increases.

  The span is the length of time in the global optimum plan, and represents how far ahead the plan will be created. In the control algorithm in this embodiment, it is one week. This is because the patterns for each day of the week are important, the calculation load is assumed not to change much between 1 day and 7 days, and weather information beyond 7 days is difficult to obtain and low accuracy. It is set in this way. Even if a failure occurs in the server of the overall optimum plan, the rules for one week ahead have already been generated. Therefore, if the rules are restored within one week, the control rules will not be blank, and the accuracy of the control rules will be improved. Just go down.

  The cycle is the update frequency of the overall optimum plan. In the control algorithm in this embodiment, it is one day. This is because there is data that is updated on a daily basis, such as weather information and EV traveling information, and for the sake of safety regarding the immediate communication load and server processing time. Further, in the present embodiment, the overall optimum plan is updated also at the time of EV connection and release. When the EV is connected / released, the power supply configuration and the actual value change greatly and discontinuously, and therefore, the plan update by event activation is performed using this as a trigger. Note that the meshes, spans, and cycles described here are merely examples, and periods other than those described here may be used.

3.6. Operation Flow of Smart House Component Device Centered on Control Algorithm FIG. 9 is a diagram illustrating an operation flow of HEMS. In this section, the relationship between the operation of smart house components (device network) such as power supplies and home appliances will be described with a focus on control algorithms. The logical configuration unit of the HEMS in this embodiment includes an overall optimal plan including a control algorithm, an analysis DB that holds actual data and control rules, an execution instruction, execution, and monitoring. Physically, the overall optimum plan and the analysis DB are mounted on the information management server in the cloud, and the execution instruction is mounted on the home server in the house. A device network consisting of a power supply and home appliances performs an operation, and a sensor network monitors the operation status.

  The analysis DB holds (stores) a performance data set used in the overall optimum plan and various parameters used for optimization calculation. Based on these data, the overall optimal plan composition unit creates a power supply and demand plan for each house. The overall optimal plan configuration unit outputs a set of control rules (control rule set) indicated by a conditional statement (for example, IF-THEN statement) to the analysis DB. The conditional expression is described, for example, by comparing time or value.

  In the analysis DB, a system administrator can register individual rules separately from the control rule set generated in the overall optimum plan. Priorities may be assigned to these rules. The analysis DB extracts a subset of rules to be executed (control rule subset) according to the priority, and transmits it to the execution instruction component, for example, every hour. Note that the communication interval between the analysis DB and the execution instruction component is not limited to one hour. Depending on the communication load of the system, communication may be performed at shorter intervals or longer intervals.

  The execution instruction configuration unit determines whether or not these satisfy the conditional expression of the control rule based on the operation status and the environment variable for each minute transmitted from the monitoring configuration unit. This determination is performed at a predetermined cycle (for example, every minute). The execution instruction configuration unit activates a control rule whose result of the conditional expression is True, and transmits the execution instruction to the execution configuration unit in, for example, the ECHONET Lite format. The frequency of the conditional expression determination and execution instruction is not limited to 1 minute, and may be set in consideration of the performance of the monitoring component, the operation status, the load of the home server or home communication, and the like.

  The operational status of the execution component is measured every minute by the monitoring component. The measurement result is transmitted to the execution instruction configuration unit.

  10 to 15 show examples of input / output of the overall optimal plan. Although there are many items, they are generally classified into parameter setting, actual value input, predicted value and scenario output, optimized plan value output, and optimization-based control rule output.

  FIG. 16 is a diagram illustrating daily processing of the overall optimal plan. FIG. 16 shows the flow of processing related to the update of the overall optimum plan once a day. The overall optimum plan receives the actual data from the execution instruction unit, the weather information from the external server and the EV traveling result, and the EV reservation from the smartphone. In the global optimum plan, optimization calculation is performed using the received information, and a control rule is generated. The control rule is transmitted to the analysis DB.

  FIG. 17 is a diagram illustrating a recalculation process of the overall optimum plan at the time of connection / release. FIG. 17 shows the flow of processing related to the update of the overall optimum plan by event activation at the time of EV connection / release. When an EV that is a part of the execution configuration unit is connected / released, a message indicating that the EV is connected / released is transmitted to the analysis DB via the monitoring configuration unit and the execution instruction configuration unit. The analysis DB calls the overall optimal plan configuration unit, and the overall optimal plan configuration unit updates the plan based on the latest data. The updated control rule is transmitted to the analysis DB.

  FIG. 18 is a diagram illustrating an hourly process of the control rule set. FIG. 18 shows the flow of processing related to the update of the control rule every hour. Performance data is transmitted from the execution instruction configuration unit to the analysis DB every hour, and conversely, the control rule subset is transmitted from the analysis DB to the execution instruction configuration unit.

  FIG. 19 is a diagram exemplifying processing for every minute of an execution instruction. FIG. 19 shows a flow of processing relating to an execution instruction for every minute. The actual value transmission from the monitoring component, the condition determination and execution instruction in the execution instruction component, and the operation execution in the execution component are repeated every minute.

  The weekly, monthly, quarterly, annual, or parameter update processing that is performed as needed is performed mainly by the system administrator periodically or appropriately in the analysis DB. The minimum power storage amount is set and updated by the user from the smartphone.

4). Module configuration of control algorithm 4.1. Module List FIG. 20 is a diagram illustrating a module list of the control algorithm. The control algorithm includes a demand module, a supply module, a power storage module, and an optimum supply and demand planning module. These are virtual reproductions of the device configuration of HEMS. Further, FIG. 20 shows an analysis DB as a configuration related to these modules.

4.2. Module Configuration Diagram FIG. 21 is a diagram illustrating a module configuration of the overall optimum plan. First, the demand module generates a demand forecast and a demand scenario based on the actual / forecast data and various parameters stored in the analysis DB. The prediction result and demand scenario of the demand module are passed to both the analysis DB and the optimum supply and demand planning module. Similarly, the supply module performs supply prediction and supply scenario generation based on the actual / forecast data and various parameters stored in the analysis DB. The prediction result and supply scenario of the supply module are passed to both the analysis DB and the optimum supply and demand plan. The optimum supply and demand planning module creates an integrated scenario, calculates an optimization plan, and generates a control rule based on various parameters stored in the analysis DB and the calculation results of the demand module and the supply module. The EV / storage battery charge / discharge plan generated by the optimal plan module, the fuel cell operation plan (start, stop, and output), and the EV, storage battery, and fuel cell control rules expressed in the conditional statements are analyzed DB Registered in In addition, among the calculation results of the optimum plan module, the respective operation plans of the system power source, the fuel cell, and the heat storage unit are returned to the supply module, and the plan of the storage amount of the EV / storage battery is returned to the power storage module. The supply module and the power storage module register these data in the analysis DB.

  In the control algorithm according to the present embodiment, the optimum supply and demand planning module calculates the planned values for the system power flow / reverse flow, the fuel cell power generation amount, and the EV / storage battery storage amount. This is because the centralized processing is performed by the optimum supply and demand planning module in consideration of the efficiency of the optimization calculation. It should be noted that different functions may be implemented by implementing different functions in the supply module and the power storage module.

4.3. Module Flow FIG. 22 is a diagram illustrating a process flow of each module. The demand module and the supply module perform prediction and scenario generation in parallel, respectively. Based on the respective prediction results and scenarios, the optimal supply and demand planning module performs integrated scenario creation, optimization calculation, and rule generation. Thereafter, among the calculation results of the optimum plan module, the respective operation plans of the system power source, the fuel cell, and the heat storage unit are returned to the supply module, and the plan of the storage amount of the EV / storage battery is returned to the power storage module.

5. Submodule configuration of control algorithm 5.1. Demand module 5.1.1. Submodule List FIG. 23 is a diagram illustrating a submodule list of demand modules. The demands to be processed by the demand module are power demand and heat demand in the house. The demand module, as the three demands to be grasped and predicted by the control algorithm of this embodiment, corresponds to the power demand of home appliances, the power demand of EV, and the heat demand of the heat storage unit, creating a home appliance power demand scenario, EV power demand It has three submodules: scenario creation and heat demand scenario creation.

5.1.2. Submodule configuration diagram FIG. 24 shows a submodule configuration of a demand module. The three sub-modules are side-by-side, each acquires actual data from the analysis DB, performs demand forecasting and demand scenario generation, and passes the demand forecast and demand scenario as the calculation results to the analysis DB and the optimal supply and demand planning module .

5.1.3. Submodule Flow FIG. 25 is a diagram illustrating a flow of a submodule of the demand module. The three submodules are side by side and may be processed in any order.

5.2. Supply module 5.2.1. Submodule List FIG. 26 shows a submodule list of supply modules. The supply module has four submodules: PV supply scenario creation, system power, fuel cell, and heat storage unit heat quantity. These are virtual reproductions of the device configuration of HEMS.

5.2.2. Submodule Configuration Diagram FIG. 27 is a diagram illustrating a submodule configuration of a supply module. The PV supply scenario creation submodule acquires actual data from the analysis DB, generates a supply scenario by supply prediction, and passes the prediction scenario as a calculation result to the analysis DB and the optimum supply and demand planning module. In the control algorithm in the present embodiment, the three submodules of the system power supply, the fuel cell, and the heat storage unit heat quantity receive the calculation result of the optimum supply and demand planning module and pass it to the analysis DB. This is because the centralized processing is performed by the optimal supply and demand planning module in consideration of the efficiency of the optimization calculation. Note that different functions may be mounted on these three submodules to perform distributed processing.

5.2.3. Submodule Flow FIG. 28 is a diagram illustrating the flow of submodules in the supply module. The PV supply scenario creation submodule is executed first. The remaining three submodules are arranged side by side, and may be processed in any order after execution of the optimum supply and demand planning module.

5.3. Power storage module 5.3.1. Submodule List FIG. 29 shows a submodule list of the power storage modules. The power storage module has two submodules: an EV power storage amount and a storage battery power storage amount. This is a virtual reproduction of the device configuration of HEMS.

5.3.2. FIG. 30 is a diagram illustrating a module configuration of the power storage module. In the control algorithm in the present embodiment, the power storage module receives the calculation result of the optimum supply and demand planning module and transfers it to the analysis DB. This is because the centralized processing is performed by the optimal supply and demand planning module in consideration of the efficiency of the optimization calculation. Note that different functions may be implemented in each submodule of the power storage module to perform distributed processing.

5.3.3. Submodule Flow FIG. 31 is a diagram illustrating a flow of submodules of the power storage module. The two submodules are side by side, and either may be processed first.

5.4. Optimal supply and demand planning module 5.4.1. Submodule List FIG. 32 is a diagram illustrating a submodule configuration of the optimum supply and demand planning module. The sub-module configuration of the optimum supply and demand planning module has three sub-modules corresponding to the necessary processing order: integrated scenario generation, optimization plan, and control rule generation.

5.4.2. FIG. 33 shows a sub-module configuration of the optimum supply and demand planning module. The three submodules are in series. First, an integrated scenario generation submodule receives a demand scenario and a supply scenario from the demand module and the supply module, respectively, and generates an integrated scenario in which these scenarios are integrated. Next, based on various parameters acquired from the integrated scenario and analysis DB, the optimization plan sub-module generates various plans as a result of the optimization calculation, and generates control rules for EV / storage battery charge / discharge plans and fuel cell operation plans. Delivered to the submodule and analysis DB. Each supply plan of the system power supply, fuel cell, and heat storage unit is transferred to the supply module, and the storage plan of the EV / storage battery is transferred to the storage module. Finally, the control rule generation submodule generates a control rule based on the EV / storage battery charge / discharge plan and the fuel cell operation plan, and registers the generated control rule in the analysis DB.

5.4.3. Submodule Flow FIG. 34 is a diagram illustrating a submodule flow of the optimum supply and demand planning module. The three submodules are serially executed in the order of the integrated scenario creation submodule, the optimization plan submodule, and the control rule generation submodule.

6). Method 6.1. Demand module 6.1.1. Home Appliance Electricity Demand Scenario Creation Submodule 6.1.1.1. Concept 6.1.1.1.1. Overview of Functions In the home appliance power demand forecasting model, first, a basic pattern (hereinafter referred to as “demand basic pattern”) is defined for different power demands depending on the day of the week, weather, etc. (hereinafter referred to as category, P). . The basic demand pattern is a representative value (average or median) calculated for each time zone (in one hour increments) within each category using the past power demand classified by category, and summarized by category. A set of values. This basic demand pattern is calculated for each house. Next, the past home appliance demand amount of the house to be predicted is matched with the basic pattern of the home by category, and the difference between the actual value of the home appliance power demand amount and the basic demand pattern is measured. Then, a model for predicting this difference is constructed. Although the seasonal factor is not an element of the above category classification, the seasonal factor affects the power demand through the temperature change, so a model with the temperature as an explanatory variable for each season is estimated. Considering the changes in the lifestyle of individual houses, it is desirable to set the basic demand pattern estimation period to several years. More specifically, it is necessary to reflect lifestyle changes such as advancement, employment, change of job, marriage, etc., and detailed hourly demand data that may be able to identify individuals is long-term and long-term for security. Since it is desirable not to hold it, the data period is preferably about 1 to 2 years.

FIG. 35 is a diagram illustrating categories of basic demand patterns. In this model, the basic demand pattern categories are classified into the following 16 types.
If it is not a holiday: Day of the week (7 types) x Presence of precipitation (2 types)
For public holidays: Precipitation (2 types)
In the present embodiment, the difference in the level of consumer electronics power demand depending on the season is not classified as a category. The difference is reflected in the forecast by estimating the model for each season, with December to March being the winter, June to September being the summer, and the rest being the interim.

  FIG. 36 is a diagram illustrating a flow of creating a home appliance power demand scenario. Regarding the prediction, first, the difference from the basic demand pattern is predicted by applying it to the relational expression of the estimation model using the season of the prediction target day and the predicted temperature. In addition, it is determined to which category the power demand of each household to be predicted is classified by day of the week / weather forecast, etc., and the basic demand pattern to be classified is specified. That is, by calculating the sum of the difference estimated by the model and the basic demand pattern, the power demand to be predicted is obtained. This estimated amount is output as a predicted value of this module. In addition, in this submodule, the fluctuation width of the predicted value is calculated from the percentile value of the error distribution of the prediction model, and the prediction value obtained by adding or subtracting the fluctuation width is output as a prediction scenario together with the expected occurrence probability.

6.1.1.1.2. Reason / Evidence This sub-module adopts a method of defining a basic demand pattern and predicting the divergence of power demand from it. The reason for this is that the power usage of houses is not uniform among households, but it is assumed that there is a pattern of power demand depending on the day of the week / presence of precipitation when viewed in individual households. (In rainy weather, stop washing, go out part-time on a specific day of the week, etc.) If there is a pattern of power demand, noise is removed from the observed value by specifying that pattern, and stable estimation A value can be obtained. Furthermore, by predicting the deviation from the basic pattern of power demand using the deviation of the predicted temperature from the average temperature, we are taking into account the speci fi c consumer electronics power demand factor, ie, the temperature change on the forecast day, to improve accuracy. .

In addition, this model adopts the method of estimating the model according to the season, and the grounds are as follows. Home appliance power demand is affected by the temperature around the house, and the relationship is assumed as follows.
Summer: The electric power demand of the house H increases as the temperature near the house H becomes larger than the average temperature (the demand is positively sensitive to the temperature change).
Winter: The electric power demand of the house H increases as the temperature near the house H becomes smaller than the average temperature (the demand is negatively sensitive to temperature changes).
Spring and Autumn: When the absolute value of the difference between the temperature near the house H and the average temperature is below a certain level, the sensitivity of the demand for home appliances to changes in temperature is so small that it can be ignored. On the other hand, when the deviation width exceeds a certain value, the power demand increases in proportion to the absolute value of the deviation.
The basis for this hypothesis is shown by the fact that the main part of household electricity demand is occupied by air conditioners. This is because the use of an air conditioner is promoted as it deviates from the temperature at which a person is likely to be comfortable, and therefore, the sign of sensitivity to temperature changes is assumed to differ between summer and winter.

  FIG. 37 is a diagram exemplifying a relationship between the temperature of home appliance power demand for each season. Even if the power demand on the vertical axis in the figure is the deviation from the basic demand pattern and the horizontal axis is replaced by the deviation from the average temperature according to the classification of the basic demand pattern, the basic demand pattern does not classify the temperature as an element. Therefore, the divergence width also has sensitivity to temperature. Therefore, in this case as well, as in FIG. 37, the summer season is a positive sensitivity, the winter season is a negative sensitivity, and the intermediate period (spring autumn season) is a bent straight line. After all, in the estimation of the summer and winter models, a linear regression model with the explanatory variables as the deviation from the average temperature for each demand pattern classification was applied, and in the spring and fall models, the same explanatory variables as in the summer and winter models were used. An LDV (Limited Dependent Variable) model is estimated.

6.1.1.2. Input / Output FIG. 38 is a diagram illustrating a list of variables / symbols used in the input / output and model of the home appliance power demand scenario creation submodule.
FIG. 39 is a diagram illustrating a variable / symbol list of the home appliance power demand scenario creation submodule.

6.1.1.3. Details of model 6.1.1.1.3.1. Summer / Winter Model FIG. 40 is a diagram illustrating the relationship between the temperature of home appliance power demand in summer and winter. The average demand amount aveE _H (H, P, t) is calculated for each of the 16 basic demand patterns described above. Note that aveE _H (H, P, t) has the same meaning as the following equation (1). In this article, it is difficult to represent a character with a bar above the symbol, so instead of adding a bar above the symbol, the prefix “ave” is used for parts other than the formula. In the mathematical formula, a character with a bar on the symbol is used as in the formula (1).
At this time, the average demand is calculated for each time mesh for a house in the vicinity of the house H (for example, the house H and other houses within a predetermined distance from the house H). P represents a category which is a type of basic pattern, and t represents a time zone in which the average demand is calculated. And the objective variable y (H, P, S, t) of the model is
It is expressed. Here, E _H (H, P, S, t) is the actual power demand value at time t of the pattern P in the seasonal housing H, and S in the formula represents the season.

On the other hand, the temperature near the house observed for each category of the basic demand pattern is classified, and the difference between the average value and the observed temperature is used as an explanatory variable of the model. In other words, the explanatory variable X (H, P, S, t) is given by the following equation as a deviation from the average temperature avem (P, S, t) for each time mesh calculated according to the basic demand pattern. .

After all, the estimation formula for the summer / winter model is
It becomes. Here, β (S) is a coefficient representing the sensitivity to changes in temperature that vary with the season, and the intercept term α (S) of the estimation formula is the average of the differences from the basic pattern of consumer electronics power demand in the season S. It is a variable that represents power demand due to factors.

6.1.1.1.3.2. Spring / Autumn Model FIG. 41 is a diagram illustrating the relationship between the temperature of home appliance power demand and the temperature in spring / autumn. As described above, an LDV model is applied as a model, but X (H, P, S, t) is applied as an explanatory variable as in the summer / winter model. As shown in FIG. 41 below, it is assumed that the sensitivity of the objective variable changes depending on the temperature level. _{Incidentally, y m (H, P,} S, t) indicates the average of the state 2 power demand y of (within the average ± T degree) (H, P, S, t).

In other words, each state in the figure is
It is expressed. This is in average temperature, within the range deviation of ± T ° C. from the basic demand patterns, sensitivity to temperature change is little, the average y _m of power demand in the state 2 (H, P, S, t) in It means that it is assumed to be constant.

  FIG. 42 is a diagram showing another example of the relationship between the temperature of home appliance power demand in spring and autumn. Now, assuming that the sensitivities to the temperature changes in state 1 and state 3 are equal, if the sign of the power demand (deviation from the average) in state 3 is reversed, FIG. 41 is transformed as shown in FIG. And the relational expression of electric power demand and temperature is estimated by estimating the temperature sensitivity of FIG.

Under the above assumption, as shown in FIG. 42, if the straight lines in state 1 and state 3 are shifted by α1 and α2, respectively, they coincide with the central straight line (FIG. 42). The degree lnL may be maximized.
In addition,
It is. A symbol i below the symbol Σ represents a state. That is, Equation (7) means that the discrete data is divided into three sections and summed.

The likelihood is explained. When the observed data belongs to state 1, the probability of occurrence of each belonging data is expressed by a probability density function. The probability density function at this time is
And the probability of occurrence for the entire state 1 data is
It becomes. Similarly, in state 3,
It becomes. here,
It is. In state 2, the sensitivity with respect to the temperature change is zero, and no change in demand is observed, but if there is a sensitivity similar to the other states and a straight line in the center of FIG. 42 is assumed, it belongs to state 2 The individual occurrence probability of the data is
It is represented by In this case as well, the “likelihood” of the entire data belonging to state 2 is
It is represented by The “likelihood” in all states is expressed by the following formula obtained by multiplying the probabilities of occurrence of all data.
This equation is the likelihood function of this model and represents the likelihood. A method of determining unknown parameters (β, α1, α2) so as to maximize this likelihood is the maximum likelihood method. In general, the likelihood function does not change in the order of magnitude even when logarithmically transformed, and therefore, a method of logarithmically converting the likelihood function for convenience and maximizing the function value is often used.

The above log likelihood function estimates the parameter to be maximized because the occurrence probability of each state can be determined if it can be determined which state (state 1, state 2, state 3) each objective variable belongs to. Can do. In the actual estimation, when T for dividing the state is known, α1 and α2 are respectively
It becomes. Actually, since T is not known, it is assumed that the data below the lower X% of the deviation (objective variable) from the basic demand pattern of the power demand amount belongs to state 2, and for other data, the data When the deviation value from the average of the temperature is positive, it is assumed that it belongs to the state 3 and when it is negative, it is assumed to belong to the state 1, and α1 and α2 are estimated from the observed values. Since X% is unknown, X% is determined by grid search so that the posterior likelihood is maximized using the parameters and states estimated to maximize the likelihood. Since the estimated parameter is used to obtain a relational expression between the temperature and the power demand, this relation is applied to the future prediction, and a future scenario is calculated.

6.1.1.1.3.3. Model Parameter Estimation Method Differentiating the likelihood function presented in Section 6.1.1.1.3.2 with the estimated parameter and setting it to zero, the following relational expression is obtained.
here,
Ni: Number of data items in state i shown in FIG. 42 (i = 1, 2, 3)
.PHI.i: Standard distribution function. The subscript below the .SIGMA. Symbol is a variable representing the state shown in FIG. 42 (i = 1, 2, 3).

  Using the above relationship, it is possible to estimate parameters so that the log likelihood converges around the maximum value using one of the EM (Expectation Maximization) algorithms, and the procedure is as follows.

Here, λ is an update width.

Step 4: The log likelihood lnL ^{(k + 1)} is calculated using the k + 1-th update parameter, and the processing from Step 2 to Step 3 is repeated until the likelihood converges.
The convergence condition may be set to a sufficiently small threshold value such that lnL ^{(k + 1)} -lnL ^(k) <threshold value. Similarly, it is possible to estimate the parameter that maximizes the likelihood using other methods, and the method to be adopted is determined in consideration of robustness, calculation time, and the like.

6.1.1.4. Creation of forecast scenario 6.1.1.1.4.1. Creation of predicted values Future predicted values are calculated by applying the relationship between temperature and power demand obtained by estimating the model equation shown in the previous section to temperature prediction. Obtain the predicted temperature in 1 hour mesh. Furthermore, considering the category and season of the basic demand pattern to which the prediction target applies, the past average value avem (H, P, S, t) corresponding to the category and season of the prediction value is calculated. Then, the deviation of the predicted value from avem (H, P, S, t) is calculated, and 6.1.1.3. By substituting into the estimation model of the clause, y ^* (H, P, S, t) as a future predicted value is obtained. Since the predicted value required as an actual scenario is not the deviation from the basic demand pattern but the demand amount itself, it is converted into demand amount data by the following formula.

Summary,
・ Predict every 1 day up to 1 week ahead.
-The basic pattern used for prediction is determined using the forecast day of the week, whether it is a national holiday, or the forecasted precipitation probability. However, the presence / absence of precipitation is “with precipitation” when the probability of precipitation is 50% or more, and “without precipitation” when the probability is less than 50%.
-Estimate a model that predicts the difference from the basic demand pattern to be applied from the season of the forecast date.
Convert the predicted temperature value near the house H in the 1-hour mesh into a deviation from the past average, and apply it to a model that estimates the difference obtained from the learning data to obtain the predicted value.

On the other hand, if 1 hour mesh temperature forecast data is not available for the future 7 days, an estimated value is obtained by the following method.
(1) Determine which of the seasons defined in this paper (summer, winter, spring / autumn) corresponds to the data to be predicted.
(2) On a daily basis, the daytime maximum and minimum temperature data for seven days ahead are released by the Japan Meteorological Agency and others. The same is true for the probability of precipitation.
there,
(A) When the predicted precipitation probability is 50% or more, it is determined that the weather on the prediction target day is “rainfall”, and when the prediction probability is less than 50%, “no rain”.
(B) With respect to the same weather and past data of the same season, the daytime maximum temperature and the minimum temperature are extracted, and the sum of squares of the respective deviations from the predicted maximum temperature and the minimum temperature is calculated.
(C) With respect to the sum of squares of the divergence calculated in (b), the data is rearranged in ascending order, and data up to about the top 10 (daily hourly increments) are acquired.
(3) Calculate the average value of household electricity demand in the past data with the highest and lowest temperatures obtained by the above procedure for each time zone (1 hour mesh), and the average value (calculated with 1 hour mesh) Is predicted data.

6.1.1.1.4.2. Setting the fluctuation range of the predicted value The difference between the calculated predicted value and the actual value is defined as the prediction error. In order to make a prediction up to a future time point one week ahead, there are seven types of errors when the type of prediction value is captured by the length of the period until the prediction date (in units of one day). As the prediction error, a past prediction value (seven types, increments of one hour) is held for each prediction time point, and the prediction error is calculated from the held data each time. It is assumed that an estimated amount related to the predicted fluctuation width is obtained from the error distribution accumulated by the above processing.

  FIG. 43 is a diagram illustrating a representative value obtained from the error distribution. The horizontal axis indicates the error, and the vertical axis indicates the probability density. FIG. 46 shows an example in which four shake widths are set. 97.5% tile points, 85% tile points, 15% tile points, and 2.5% tile points are extracted (or estimated by linear interpolation) from the past prediction error data based on the magnitude of the error. Set as the fluctuation width of the predicted value.

6.1.1.4.2.3. Setting of prediction scenario When setting a prediction scenario using the set fluctuation width of the above-mentioned prediction value, “prediction value”, “prediction value + 85% tile point”, “prediction value + 15% tile point”, “prediction” Value + 97.5% tile point ”and“ Predicted value + 2.5% tile point ”are output as scenarios with occurrence probabilities of 50%, 20%, 20%, 5%, and 5%, respectively.

6.1.2. EV power demand scenario creation submodule 6.1.2.1. Concept 6.1.2.1.1. Overview of Function FIG. 44 is a diagram for explaining an EV power demand scenario creation flow. In this sub-module, the electric power demand accompanying EV use is predicted, and an EV electric power demand scenario considering the predicted value and its error is created. However, it is difficult to obtain observation variables that enable prediction of future EV usage. Therefore, the prediction of EV use is a scenario in which the resident's use reservation is plausible from the viewpoint of occurrence probability, and the scenario is set as a predicted value of this module. For this scenario, the usage reservation information is converted into electric power according to the usage reservation time zone and its length, and is set as an output scenario. On the other hand, a scenario output other than the main scenario (which is likely from the viewpoint of occurrence probability) is obtained by sampling at random from the EV power demand record in order to reflect the experience distribution.

6.1.2.1.2. Reason / Evidence Although it is difficult to predict, sampling is performed from actual EV power demand performance data, and the scenario creation reflecting the actual usage situation is realized by reflecting the experience distribution. The use reservation is a result of the intention display of the user himself, and it is reasonable to use this as the main scenario.

6.1.2.2. Input / Output FIG. 45 is a diagram illustrating EV power demand scenario creation submodule input / output.
FIG. 46 is a diagram illustrating an EV power demand scenario creation submodule variable / symbol list.

6.1.2.3. Details of the model EV use reservation information is EV use time zone information, and this information needs to be converted into electric power demand. In this submodule, there is no statistical model, and there is only logic for converting EV usage reservations into power demand.

6.1.2.4. Creation of forecast scenario 6.1.1.2.4.1. Creation of Predicted Value As described above, a plausible prediction is user input information called EV use reservation. Therefore, the EV usage reservation converted into the power demand is the predicted value of the submodule.

6.1.2.4.2. Setting of fluctuation range of predicted value Scenarios other than the predicted value are random sampling from past actual values. Therefore, no specific prediction value fluctuation is set.

6.1.4.2.3. Prediction scenario setting A prediction scenario is created for seven days (prediction period) by random sampling from data for the past year stored in the EV power demand record DB. The procedure will be described in detail below.

Step 1:
Calculate the day of the week for the forecast date. (Assumed to be day x below)
Step 2:
The past performance data relating to the EV power demand is acquired, the demands are summed daily, and the data included in the top 1% having the highest total value is excluded from the acquired data.
Step 3:
The data of x day of the acquired past data and its data date are extracted.
Step 4:
Data other than the data that can be acquired continuously (not missing) for 7 days including the data date are deleted from the extracted data in step 3.
Step 5:
The remaining data is sorted in ascending order of date, and the number of remaining data is counted.
Step 6:
Uniform random numbers are generated for the number of counts in step 5, and random numbers are assigned to each data in the order of data generation.
Step 7:
Assuming that n-1 scenarios (n-1 needs to be equal to or smaller than the count number in step 5) are acquired, n-1 data are extracted in ascending order of the random numbers given in step 6.
Step 8:
Including the data date of the data extracted in step 7, data for 7 days after each data date is acquired, and a set of data continuous for 7 days starting from day x is collected as a unit scenario. At this point, n-1 scenarios are created.
Step 9:
The scenario (n-1) summarized in step 8 is output.

  Next, regarding the setting of the occurrence probability of these scenarios, there are a total of n scenarios with n-1 sub-scenarios and one main scenario. Since the main scenario is a plausible scenario (it is assumed that the occurrence probability is the highest), a high occurrence probability should be given to the scenario.

The specific occurrence probability to be given assumes an occurrence probability of 80% to 90% at this stage. If the setting is 80%, the remaining 20% is allocated to the other n-1 scenarios. On the other hand, since these scenarios are obtained by random sampling, it is reasonable to set each occurrence probability equally. Therefore, the occurrence probability is set as (100-80)% that is equally distributed to these n-1 scenarios. That is, the occurrence probability of scenarios other than the main scenario is given by the following formula.

6.1.3. Heat demand scenario creation submodule 6.1.3.1. Concept 6.1.3.1.1. Overview of Function FIG. 47 is a diagram for explaining a flow of creating a heat demand scenario. It is assumed that heat demand has a strong correlation with changes in temperature, season, and day of the week. In addition, it is assumed that heat use follows a certain pattern in each house for each season and day of the week (hereinafter referred to as category). Therefore, the prediction model of this submodule is a prediction model that uses past actual heat demand data according to the above hypothesis. Specifically, the past daily heat demand performance data is classified into each category. Next, it is determined which category the prediction target date belongs to. Moreover, the daytime maximum temperature and minimum temperature data of the prediction target date are acquired. On the other hand, regarding the category for which the predicted date is determined, the daytime maximum temperature and the minimum temperature on the data date of the past actual demand data of the same category are calculated. Then, from the past performance data of the category corresponding to the forecast date, multiple past performance data with the daytime maximum and minimum temperature forecasts on the forecast date and the daytime maximum and minimum temperatures on the data date are extracted, and the representative The value is applied as forecast demand data. Whether or not the temperature is close is determined by calculating the square of the difference for each of the lowest temperature and the highest temperature, and using the value obtained by calculating the sum of each difference as a reference.

6.1.3.1.2. Reason / Evidence Heat demand correlates with temperature, such as through floor heating. Also, the use of hot water during bathing differs between summer and winter. Furthermore, the bathing time may differ on holidays and weekdays. In order to deal with these, it is necessary to consider factors such as day of the week, season, and temperature in the prediction. Therefore, in this submodule, past performance data is classified into categories of combinations of seasons and days of the week, the representative value of demand is extracted from the categories corresponding to the forecast date using the temperature as the selection criterion, and the forecast value is calculated. The calculation method is adopted.

6.1.3.2. Input / Output FIG. 48 is a diagram illustrating input / output of the heat demand scenario creation submodule.

6.1.3.3. Model details The prediction of this sub-module is made according to the following procedure.
Step 1:
Classify the past heat demand performance data into the following 24 categories If it is not a holiday: Day of the week (7 categories) x Season (3 categories)
For holidays: Season (3 categories)
Winter: December to March Summer: June to September Interim: April, May, October, November Step 2:
It is determined to which category defined above the data to be predicted corresponds to the prediction time point.
Step 3:
On a daily basis, the daytime maximum and minimum temperature data for seven days ahead are published by the Japan Meteorological Agency and others. there,
(1) With respect to past data of the same category, each daytime maximum temperature is taken out, and the square of the difference from the predicted daytime maximum temperature is calculated. The same calculation is applied to the minimum temperature.
(2) The sum of the squares of the minimum temperature and the maximum temperature calculated in (1) is calculated daily, and a plurality of higher-order data (in daily hour increments) with a small difference are acquired.
(3) The average value (in one hour increments) of the heat demands of the acquired plurality of upper data is calculated as a representative value in each mesh, and the representative value is used as prediction data.

6.1.3.4. Creation of forecast scenario 6.1.1.3.4.1. Creation of predicted values In this submodule, sampling by category is performed from past performance data, so the representative values of past data obtained in the model details are the predicted values themselves.

6.1.3.2.2. Setting the fluctuation range of the predicted value The difference between the calculated predicted value and the actual value is defined as the prediction error. In order to make a prediction up to a future time point one week ahead, there are seven types of errors when the type of prediction value is captured by the length of the period until the prediction date (in units of one day). As the prediction error, a past prediction value (seven types, increments of one hour) is held for each prediction time point, and the prediction error is calculated from the held data each time. It is assumed that an estimated amount related to the predicted fluctuation width is obtained from the error distribution accumulated by the above processing.

  FIG. 49 is a diagram illustrating a representative value obtained from the error distribution. The horizontal axis indicates the error, and the vertical axis indicates the probability density. FIG. 49 shows an example in which four swing widths are set. 97.5% tile points, 85% tile points, 15% tile points, and 2.5% tile points are extracted (or estimated by linear interpolation) from the past prediction error data based on the magnitude of the error. Set as the fluctuation width of the predicted value.

6.1.4.3. Setting of prediction scenario When setting a prediction scenario using the set fluctuation width of the above-mentioned prediction value, “prediction value”, “prediction value + 85% tile point”, “prediction value + 15% tile point”, “prediction” Value + 97.5% tile point ”and“ Predicted value + 2.5% tile point ”are output as scenarios with occurrence probabilities of 50%, 20%, 20%, 5%, and 5%, respectively.

6.2. Supply module 6.2.1. PV supply scenario creation submodule 6.2.1.1. Concept 6.2.1.1.1. Function Overview FIG. 50 is a diagram for explaining a PV supply amount scenario creation flow. In this module, a non-linear regression model representing the relationship with photovoltaic power generation (hereinafter referred to as PV) is estimated using the amount of solar radiation and temperature in the past normal years. The installation angle factor can be explicitly incorporated into the model, but is not used by default. Then, using the estimated model parameters, it is applied to forecast data (weather and temperature) for 7 days (predicted time increment is 1 hour) and solar radiation (previous results are used as forecast values), and future supply of PV Predict the amount. The prediction model is a non-linear regression model with factors of a power generation efficiency function and a solar radiation amount decrease coefficient due to weather (a coefficient of a dummy variable indicating whether the weather is clear). Further, a percentile value of error (for example, 2.5%, 15%, 85%, 97.5%) is obtained from the past prediction error distribution, and a prediction scenario is created together with the expected value.

6.2.1.1.2. Reason / Evidence It is known that the amount of solar radiation draws the same unimodal shape in a 24-hour cycle if the region and the weather are the same. In addition, power generation efficiency and temperature are in a substantially linear relationship. Therefore, it is considered that power generation forecasting can be modeled by the product of a linear function of power generation efficiency with temperature as a variable and the amount of solar radiation, and daily performance data can be applied as an alternative to the predicted value of solar radiation. . On the other hand, considering the solar radiation loss coefficient due to weather (dummy variable coefficient for clear weather), the estimation model is expressed by the product of the solar radiation function including the solar radiation decrease coefficient and power generation efficiency (non-linear). A nonlinear regression formula is adopted as the model formula.

6.2.1.2. Input / Output FIG. 51 is a diagram illustrating input / output of the PV supply amount scenario creation module.
FIG. 52 is a diagram illustrating a variable / symbol list of the PV supply amount scenario creation module.

6.2.1.3. Details of Model FIG. 53 is a diagram illustrating the relationship between solar radiation and power generation efficiency. Assume that the PV supply is expressed as the product of solar radiation and PV power generation efficiency.

FIG. 54 is a diagram illustrating the relationship between PV power generation efficiency and temperature. The PV power generation efficiency is a product of a linear function (straight line) related to the temperature and an installation angle coefficient.

FIG. 55 is a diagram illustrating the amount of solar radiation due to the weather. In the estimation model, the amount of solar radiation uses daily (24-hour cycle) performance data corresponding to the predicted area. Since it is reasonable to assume that the ratio is lower during cloudy weather and rainy weather than during sunny weather, a solar radiation amount reduction coefficient is introduced. Therefore, the amount of solar radiation applied to the forecast is the amount of solar radiation in a multi-year that is publicized by a predetermined external organization or business operator, and the decrease in the amount of solar radiation due to the weather is reduced in the model according to the actual situation. The solar radiation amount and the solar radiation amount reduction coefficient are represented by the following symbols.
And the amount of solar radiation in cloudy and rainy weather is
It is represented by

As described above, the estimated nonlinear regression model is set as the following regression equation (40) using the defined variables.
here,
It is. The model uses solar radiation l {H, t} and temperature θ (t) as inputs, temperature coefficient γ _S θ (H) of power generation efficiency, temperature constant γ _S0 (H) of power generation efficiency, and reduction coefficient γ _{lcloudy of} solar radiation _efficiency. Is estimated. The objective variable is the PV supply amount s (H, t). The above regression equation is initially set as a regression equation that does not include the installation angle coefficient because the installation angle coefficient and the power generation efficiency parameter cannot be determined for estimation. Therefore, in this case, the power generation efficiency parameters γ _S θ (H) and γ _S0 (H) are estimated as values including the effect of the installation angle coefficient.

As a model formula, the above nonlinear regression model is assumed.
It becomes. In this case, the model formula is a linear model for θ (t) l {H, t} and l {H, t}, and prediction is performed using this formula.

6.2.1.4. Creation of forecast scenario 6.2.2.1.4.1. Creation of predicted values Applying the above model formula to past data and estimating the parameters, the relationship between temperature, solar radiation, weather, and PV supply, future weather forecast, predicted temperature, and past solar radiation Applied to, and calculate the predicted value in the future. The prediction is carried out for the PV supply amount for the future 7 days (in increments of 1 hour) including the prediction date.

6.2.2.1.4.2. Setting the fluctuation range of the predicted value The difference between the calculated predicted value and the actual value is defined as the prediction error. In order to make a prediction up to a future time point one week ahead, there are seven types of errors when the type of prediction value is captured by the length of the period until the prediction date (in units of one day). As the prediction error, a past prediction value (seven types, increments of one hour) is held for each prediction time point, and the prediction error is calculated from the held data each time. It is assumed that an estimated amount related to the predicted fluctuation width is obtained from the error distribution accumulated by the above processing.

  FIG. 56 is a diagram illustrating representative values obtained from the error distribution. FIG. 56 shows an example in which four runout widths are set. 97.5% tile points, 85% tile points, 15% tile points, and 2.5% tile points are extracted (or estimated by linear interpolation) from the past prediction error data based on the magnitude of the error. Set as the fluctuation width of the predicted value.

6.2.4.1.4.3. Setting of prediction scenario When setting a prediction scenario using the set fluctuation width of the above-mentioned prediction value, “prediction value”, “prediction value + 85% tile point”, “prediction value + 15% tile point”, “prediction” Value + 97.5% tile point ”and“ Predicted value + 2.5% tile point ”are output as scenarios with occurrence probabilities of 50%, 20%, 20%, 5%, and 5%, respectively.

6.2.2. System power supply submodule 6.2.2.1. Concept 6.2.2.1.1. Overview of functions Conceptually a sub-module that virtually reproduces the equipment configuration of HEMS, and stores the optimal power flow / reverse power flow plan of the grid power generated by the optimal supply and demand planning module in the analysis DB. It is a submodule.

6.2.2.1.2. Reason / Evidence The control algorithm according to the present embodiment has a simple function because the centralized processing is performed by the optimum supply and demand planning module in consideration of the efficiency of the optimization calculation. Note that different functions may be mounted on the system power supply submodule to perform distributed processing.

6.2.2.2. Input / Output FIG. 57 is a diagram illustrating input / output in the system power supply submodule.

6.2.3. Fuel cell submodule 6.2.3.1. Concept 6.2.2.3.1.1. Overview of Function The fuel cell sub-module is a sub-module that virtually reproduces the HEMS device configuration conceptually. Specifically, the fuel cell submodule is a submodule for storing an optimal operation plan (startup, stoppage, output) of the fuel cell generated by the optimal supply and demand planning module in the analysis DB.

6.2.2.3.1.2. Reason / Evidence The control algorithm according to the present embodiment has a simple function because the centralized processing is performed by the optimum supply and demand planning module in consideration of the efficiency of the optimization calculation. Note that different functions may be implemented in the fuel cell submodule to perform distributed processing.

6.2.3.2. Input / Output FIG. 58 is a diagram illustrating input / output in the fuel cell submodule.

6.2.4. Thermal storage unit calorie sub-module 6.2.4.1. Concept 6.2.4.1.1.1. Overview of Function The heat storage unit heat quantity submodule is a submodule that virtually reproduces the HEMS device configuration. Specifically, the heat storage unit heat amount submodule is a submodule for storing the optimum heat storage amount plan of the heat storage unit generated by the optimum supply and demand planning module in the analysis DB.

6.2.4.1.2. Reason / Evidence The control algorithm according to the present embodiment has a simple function because the centralized processing is performed by the optimum supply and demand planning module in consideration of the efficiency of the optimization calculation. In addition, a different function may be mounted on the heat storage unit heat quantity submodule to perform the distributed processing.

6.2.4.2. Input / Output FIG. 59 is a diagram illustrating input / output in the heat storage unit heat quantity submodule.

6.3. Power storage module 6.3.1. EV storage amount submodule 6.3.1.1. Concept 6.3.1.1.1. Overview of Functions The power storage module is a sub-module that conceptually reproduces the HEMS device configuration. Specifically, the power storage module is a sub-module for storing an optimal storage amount plan of EV generated by the optimal supply and demand planning module in the analysis DB.

6.3.1.1.2. Reason / Evidence The control algorithm according to the present embodiment has a simple function because the centralized processing is performed by the optimum supply and demand planning module in consideration of the efficiency of the optimization calculation. Note that different functions may be implemented in the EV storage amount submodule to perform distributed processing.

6.3.1.2. Input / Output FIG. 60 is a diagram illustrating input / output in the EV storage amount submodule.

6.3.2. Storage battery storage amount sub-module 6.3.2.1. Concept 6.3.2.1.1. Overview of Functions The storage battery storage amount submodule is a submodule that virtually reproduces the HEMS device configuration. Specifically, the storage battery storage amount submodule is a submodule for storing the optimal storage amount plan of the storage battery generated by the optimal supply and demand planning module in the analysis DB.
6.3.2.1.2. Reason / Evidence The control algorithm according to the present embodiment has a simple function because the centralized processing is performed by the optimum supply and demand planning module in consideration of the efficiency of the optimization calculation. Different functions may be implemented in the storage battery storage amount submodule to perform distributed processing.

6.3.2.2. Input / Output FIG. 61 is a diagram illustrating input / output in the storage battery storage amount submodule.

6.4. Optimal supply and demand planning module 6.4.1. Integrated scenario generation submodule 6.4.1.1. Concept 6.4.1.1.1. Function Overview FIG. 62 is a diagram for explaining the flow of the integrated scenario creation submodule. In this submodule, the scenario created from the consumer electronics, EV, and heat demand scenario creation submodule in the demand module and the scenario created from the PV supply submodule in the supply module are integrated together with their occurrence probabilities. The function to pass the scenario to the optimization module. Here, the integration means that the respective submodule scenarios created independently are combined with each other in consideration of the correlation structure. The integrated scenario in this paper is the combined scenario, and its occurrence probability is estimated in this submodule.

6.4.1.1.2. Reason / Evidence The optimization module probability programming assumes a random variable that follows a discrete distribution. Therefore, it is necessary to express the distribution of predicted values by generating a plurality of sets (scenarios) of values and their occurrence probabilities p. However, since the prediction sub-modules (home appliance power demand prediction, heat demand prediction, EV power demand prediction, PV supply amount prediction) each perform prediction independently, the occurrence probability is calculated for each prediction combination. It is necessary to calculate separately. The reason is that the scenario of each prediction model is not independent through temperature and weather except for EV power demand prediction. If each scenario is assumed to be independent, a combination of scenarios with low feasibility is generated as an integrated scenario, giving the same probability of occurrence as other scenarios with high feasibility. End up.

6.4.1.2. Input / Output FIG. 63 is a diagram illustrating input / output of the integrated scenario creation submodule.
FIG. 64 is a diagram illustrating a list of variables and symbols of the integrated scenario creation submodule.

6.4.1.3. Details of the model Details of the model of this submodule are as follows. First, with respect to each of the PV supply amount prediction, the home appliance power demand prediction, and the heat demand prediction, time series data of errors is acquired from a database storing errors. Then, an error correlation matrix is estimated using them. The correlation coefficient is estimated by rank correlation because the error distribution may be a non-normal distribution. On the other hand, it is assumed that each prediction model has a common factor that is not observed. This is because, in fact, information such as temperature and weather is adopted as an explanatory variable, and the error is assumed to reflect the lifestyle of the house that is the object of prediction.

  Although this common factor is not directly observed, it is assumed here that it follows a normal distribution for convenience. In fact, it is not necessarily a normal distribution, but by adopting a multidimensional normal distribution function for the error distribution of these prediction models, compared to the case where the predicted values are assumed to be independent, The probability of occurrence is obtained.

  A three-dimensional normal density function (dispersion is 1) is set using the obtained correlation matrix. Since the scenario of each prediction model is obtained as the percentile value of the error of each model, they are applied to the standard normal distribution function and converted to values when it is assumed to follow the standard normal distribution. Specifically, the numerical value obtained as the 75% tile point can be calculated by giving 75% to the inverse function of the standard normal distribution function, and the value is about 0.6745. In the case of a 25% tile point, it is -0.6745.

  FIG. 65 is a diagram for explaining a flow for generating a co-occurrence probability of a scenario assuming a multidimensional normal density function. By substituting these converted numerical values into the above three-dimensional normal density function, a scenario co-occurrence probability can be obtained.

  Through the above processing, the scenario co-occurrence probability obtained from each of the PV supply amount prediction model, the home appliance power demand prediction model, and the heat demand prediction model was calculated. In this module, it is necessary to calculate the co-occurrence probability of the integrated scenario when the EV power demand scenario is combined with this value. Here, since it is assumed that the EV power demand is determined independently of other models, the occurrence probability of the scenario is calculated based on the scenario of the above-mentioned three models (PV supply amount, home appliance power demand amount, heat demand amount). By multiplying the co-occurrence probability, it is possible to calculate the co-occurrence probability of the scenario of 4 models (PV supply amount, home appliance power demand amount, heat demand amount, EV power demand amount) that also considers the EV power demand amount. Become.

  FIG. 66 is an image diagram when there are five scenarios of the PV supply amount prediction model, the home appliance power demand prediction model, and the heat demand prediction model. Since there are 125 combinations of the three model scenarios, the scenarios of the EV power demand prediction model are combined with the 125 scenarios to form a combination of integrated scenarios. The method for calculating the occurrence probability is as described above.

6.4.1.4. Generation of integrated scenario By following the method described in the model details in the previous section, the scenario co-occurrence probability is obtained for all combinations of scenarios related to the home appliance power demand model, PV supply volume model, heat demand model, and EV power demand model. In consideration of the fact that the integration scenarios calculated taking into account concurrency include those that are extremely unlikely to occur and the actual load on the optimization module, this sub-module is a scenario with a high probability of occurrence. Select the final integration scenario. Then, each occurrence probability of the integrated scenario is a value normalized so that the total occurrence probability of the selected scenario becomes 100%. The normalized occurrence probability is calculated by the following formula.

  FIG. 67 is a diagram illustrating generation of an integrated scenario. FIG. 67 shows an example when m scenarios are selected.

6.4.2. Optimization plan submodule 6.4.2.1. Concept 6.4.2.1.1. Overview of functions In this submodule, integration scenarios in units of mesh time (1 hour) for one future span (7 days) generated from household electrical power demand forecast, EV power demand forecast, heat demand forecast and PV supply forecast Is used as an input to calculate an optimal control method for the device during the period. For optimization, stochastic mixed integer programming, which is one of mathematical programming methods, is used. In probabilistic mixed integer programming, for example, a control plan is calculated that takes into account the risk when the predicted value deviates, while taking into account control constraints such as battery charging and discharging not occurring simultaneously. . The index value of optimality is set by adjusting the input item according to the specified optimization mode. For example, in the “economic mode”, the total grid power and fuel (gas) usage fee in one span are used as index values, and a control plan with a focus on reducing the fee is output. The output from the submodule is the control of each device (ON / OFF or the output level of operation or the amount of stored battery) in units of mesh periods for one future span equal to the scenario. In addition to the standard power storage level, the storage battery has an upper storage level that corresponds to risk-avoidance control, and a lower storage level that is more risk-tolerant and places more weight on optimization mode settings. Output. In this system, it is assumed that either one stationary storage battery or EV is installed in one house, but the algorithm of this submodule theoretically has multiple stationary storage batteries and EVs installed. Even in the case where each device is controlled, it can be expanded as long as each device can be controlled independently. However, for example, when there are control restrictions such that two or more EVs cannot be discharged and charged at the same time, it is necessary to consider adding a model formula.

6.4.2.1.2. Reason / Evidence In the control of a HEMS device including a storage battery as in the present system, the viewpoint of controlling the device based on future prediction is important. For example, the home appliance power demand and the PV power generation amount that are the basis of the optimization calculation are predicted values in the future, and the results as such are not always realized. Therefore, how much blur is likely to occur in the future is expressed as a scenario with a plurality of predicted value transition cases and their occurrence probabilities as scenarios. In order to calculate the control considering the magnitude of blur for each transition of the predicted value expressed in the scenario as a risk and minimizing the future risk, an optimization method that can take uncertainty into consideration is necessary. Therefore, stochastic programming is adopted. Furthermore, there are two alternative constraints on battery control, such as not charging and discharging at the same time. To handle this, an optimization problem framework that includes discrete variables is required. Therefore, stochastic mixed integer programming is adopted.

6.4.2.2. Input / output Summarizes the input / output of this sub-module and the symbols used for its definition.
6.4.2.2.1. Input FIG. 68 is a diagram illustrating items of the integration scenario.
69 to 72 are diagrams illustrating parameter items. Input includes a scenario generated by the integrated scenario creation submodule and parameter items that are constants for optimization processing.

6.4.2.2.2. Output FIG. 73 is a diagram illustrating an output of this submodule.

6.4.2.3. Model Details The elements that make up the optimal model used in this submodule include:
1) Variables 2) Restrictions 3) There are three objective functions. The variable is an amount that represents device control (ON / OFF, the amount of charge / discharge, etc.), the amount of change in electric power or heat, and the like, and is a target whose value is calculated by optimization processing.

The relationship between these quantities expressed by mathematical expressions is a constraint condition. The constraint must be defined by a plurality of linear equalities or inequalities. In general, when N variables are expressed as x1, x2,..., XN, a linear equality or inequality is obtained by using N constants a1, a2,. Given a constant b,
It is determined. The objective function is a mathematical formula for calculating an index for minimizing or maximizing the optimization, for example, the total cost of electric power and gas, and becomes a linear formula by giving a coefficient of each variable as in the constraint condition. It is determined as follows.

6.4.3.2.3.1. Variables FIGS. 74 to 76 are diagrams illustrating a list of variables used in the optimization model. The optimization variables are device control (ON / OFF, discharge amount, etc.), power and heat for each house (H), for each period (t = 1, 2,..., T) corresponding to the planned mesh. It corresponds to the amount of change.

6.4.4.2.3.2. Restriction 6.4.2.3.2.1. Constraint Condition Representing Balance between Electricity Supply / Demand and Heat Supply / Demand FIG. 77 is a diagram illustrating an example of the configuration of equipment considered in optimization. FIG. 77 shows the relationship between variables and devices considered in optimization and each variable.

The following constraints are derived from the condition where the power coming out of the circuit in the system and the power coming in are balanced. In addition, since the electric home appliance power demand E _Hi (H, t) and the PV battery supply amount s _i (H, t) are different for each scenario i, the recourse variables e _i ⁺ (H, t) and e _i ⁻ (H, The difference (power surplus and power shortage) is expressed using t). This difference is incorporated into the optimization objective function as a risk and minimized.
In addition, there is an upper limit to the amount of power purchased from the grid and the reverse power flow, and the following constraint equation is considered.
Similarly, the following constraints are derived from the condition where the heat coming out of the system and the incoming heat are balanced.
Again, since the heat demand Q _i (H, t) for each scenario is different, recourse variable ω _i ⁺ (H, t) and ω _i ^- (H, t) the difference (heat surplus and heat with Deficiency). As in the case of power, this difference is incorporated into the optimization objective function as a risk and minimized.

6.4.2.3.2.2. Constraints on FC power generation amount The relationship between the FC power generation amount p _FC (H, t) and the amount of gas fuel to be input g (H, t) is linear. If the proportionality coefficient is C _FC ^p (H) and the intercept term is C _FC ^p0 (H), the relationship between the power generation amount of FC and the input fuel is expressed by the following linear expression.

Assuming that the amount of power generation per unit time of FC has a lower limit p _FC ^min (H, t) and an upper limit p _FC ^max (H, t), the following constraint conditions are introduced.
If the FC operation is scheduled to be canceled in the future period t due to maintenance, etc., the upper limit power generation amount p _FC ^max (H, t) is set to 0, so that Operation interruption can be taken into account. Further, δ _FC (H, t) is a variable for distinguishing the start and stop states of the FC storage battery, and is defined as a discrete variable of 0-1 that becomes 1 if it is started during the period t and 0 if it is stopped. Variable. Note that modeling was performed assuming that FC start and stop are not repeated within one unit mesh time. This is because starting power is consumed when starting from a stop, and therefore, it is assumed that the effect is limited even if modeling is performed assuming fine starting and stopping within the mesh.

Next, a constraint condition related to FC start or stop is defined. If the variable δ _FC ^ON (H, t) is a 0-1 variable that is 1 when the FC transitions from the stopped state to the activated state at the start of the period t, and 0 otherwise, it is activated and stopped when activated. Since δ _FC (H, t) = 1 and δ _FC (H, t−1) = 0, the following constraint condition is satisfied.

Conversely, when the fuel cell is stopped, the variable δ _FC ^OF (H, t) is set to 1 when the FC transitions from the start state to the stop state at the start of the period t, and 0 otherwise. If the variable is −1, the following constraint condition is satisfied.
The variables δ _FC ^ON (H, t) and δ _FC ^OF (H, t) related to the start and stop are incorporated in the objective function, and optimization is performed in consideration of the number of start and stop.

6.4.2.3.2.3. Restriction condition regarding EV storage battery First, it is considered that there is a lower limit and an upper limit in the amount of electricity stored in the EV storage battery at the beginning of the period t, and the restriction condition is set as follows.
The lower limit value V _min (H, t) and the upper limit value V _max (H, t) can reflect the setting value of the user for each time in addition to the case where it is determined by the specifications of the device. If the variables representing the charge amount and the discharge amount are v ⁺ (H, t) and v ⁻ (H, t), respectively, the constraint conditions indicating that charging and discharging cannot be performed simultaneously are as follows.
Note that v _max (H, t) is set to 0 when there is an EV operation stop in the future period t due to maintenance or the like. Further, δ ⁺ (H, t) and δ ⁻ (H, t) are variables defined as 0-1 discrete variables for distinguishing the charged state and the discharged state of the storage battery. Not 1. Therefore, there is a constraint that the sum of δ ⁺ (H, t) and δ ⁻ (H, t) is 1 or less.

Further, for each scenario i, the following is defined as a constraint condition indicating that neither charging nor discharging is possible when the EV is not connected.
Note that δ _EVi (H, t) is a constant that is 1 if EV is connected and can be discharged and discharged in scenario i, and is 0 otherwise, and δ _i ⁺ (H, t) and δ _i ^- (H, t) is a discrete variable of 0-1 that distinguishes the charged state and discharged state of the storage battery introduced for each scenario i.

That is, when EV is connected in scenario i, δ _EVi (H, t) = 1, so variables v ⁺ (H, t) and v ⁻ (H, t) are 0, and the constraint condition is ,
Thus, it can be expressed that either charging or discharging is performed.

On the other hand, when EV is not connected in scenario i, the sum of variables δ _i ⁺ (H, t) and δ _i ⁻ (H, t) becomes 0 from δ _EVi (H, t) = 0. As a result, the constraint condition is
It becomes. Therefore, both v ⁺ (H, t) −v _i ⁺ (H, t) and v ⁻ (H, t) −v _i ⁻ (H, t) are 0, and the battery is neither charged nor discharged in scenario i. It was expressed.

Next, a constraint condition representing the transition of the amount of electricity stored in the EV storage battery is defined. The storage amount V (H, t) of the EV storage battery at the beginning of the period t changes to V (H, t + 1) at the beginning of the next period t + 1 according to the amount of power reduction due to charging / discharging or EV use. It is expressed by the constraint equation.
Among the above constraints, η _V ^C (H) and η _V ^D (H) are the charging efficiency and discharging efficiency of the EV storage battery, respectively. Moreover, since the use of EV is different for each scenario i, recourse variable g _i ⁺ (H, t) and g _i ^- expresses the (H, t) the difference (power surplus and power shortage) using. If the initial storage capacity of the EV storage battery at the start of the processing span period is set to V ₀ , the following constraint condition is satisfied.

Next, a constraint condition for expressing the number of times that charging and discharging of the EV storage battery are switched is introduced. The variable δ _V ^C (H, t) is a 0-1 variable that is 1 when the EV storage battery transitions from the discharged state to the charged state in the period t, and 0 otherwise. Similarly, the variable δ _V ^D (H, t ) Is a 0-1 variable that is 1 when the EV storage battery transitions from the charged state to the discharged state during period t, and 0 otherwise. When the state changes from the discharging state to the charging state, the variables representing charging are δ ⁺ (H, t) = 1 and δ ⁺ (H, t−1) = 0, so the variable δ _V ^C (H, t ) Satisfies the following constraints.

Similarly, when the EV storage battery transitions from the charged state to the discharged state, the variables representing discharge are δ ⁻ (H, t) = 1 and δ ⁻ (H, t−1) = 0, so that the variable δ _V ^D (H, t) satisfies the following constraint condition.
Variables δ _V ^C (H, t) and δ _V ^D (H, t) relating to activation and deactivation are incorporated into the objective function, and optimization is performed in consideration of the number of activations and deactivations.

6.4.2.3.2.3. Constraints on stationary storage battery As with EV storage batteries, the storage amount R (H, t) of the stationary storage battery at the beginning of period t is considered to have a lower limit and an upper limit, and the constraint conditions are set as follows: .
The lower limit value R _min (H, t) and the upper limit value R _max (H, t) can reflect the user's set value for each time in addition to the case where the lower limit value R _min (H, t) is determined by the specifications of the device.

If the variables representing the discharge amount and the charge amount are r ⁺ (H, t) and r ⁻ (H, t), respectively, the constraint conditions indicating that charging and discharging cannot be performed simultaneously are as follows.
It should be noted that r _max (H, t) is set to 0 when the storage battery operation is stopped in a future period t due to maintenance or the like. Further, δ _R ⁺ (H, t) and δ _R ⁻ (H, t) are variables defined as discrete variables of 0-1 that distinguish between the charged state and the discharged state of the stationary storage battery. The sum is not 1 at the same time, and the sum is expressed as 1 or less.

Next, a constraint condition representing the transition of the amount of power stored in the stationary storage battery is defined. Expressing that the charged amount V (H, t) of the stationary storage battery at the beginning of the period t changes to R (H, t + 1) at the beginning of the next period t + 1 by charging / discharging by the following constraint equation: Yes.
Among the above constraints, η _B ^C (H) and η _B ^D (H) are the charging efficiency and discharging efficiency of the stationary storage battery, respectively. Note that if the initial storage capacity of the stationary storage battery at the start of the processing span period is R ₀ , the following constraints are imposed.

Next, a constraint condition for expressing the number of times the charging and discharging of the stationary storage battery are switched is introduced. The variable δ _R ^C (H, t) is a 0-1 variable that is 1 when the stationary storage battery transitions from the discharged state to the charged state in the period t, and 0 otherwise. Similarly, the variable δ _R ^D (H, t) Let t) be a 0-1 variable that is 1 when the stationary storage battery transitions from the charged state to the discharged state during period t, and 0 otherwise. When transitioning from the discharged state to the charged state, the variables representing charging are δ _R ⁺ (H, t) = 1 and δ _R ⁺ (H, t−1) = 0, so the variable δ _V ^C (H , T) satisfies the following constraint condition.

Similarly, when the stationary storage battery transitions from the charged state to the discharged state, the variables representing the discharge are δ _R ⁻ (H, t) = 1 and δ _R ⁻ (H, t−1) = 0. The variable δ _V ^D (H, t) satisfies the following constraint condition.
Variables δ _R ^C (H, t) and δ _R ^D (H, t) relating to activation and deactivation are incorporated into the objective function and are optimized in consideration of the number of activations and deactivations.

6.4.2.3.2.5. Constraints on the heat storage unit The relationship between the heat generation amount p _FC ⁱⁿ (H, t) of _FC and the amount of gas fuel g (H, t) to be input is linear. If the proportionality coefficient is C _FC ^q (H) and the intercept term is C _FC ^q0 (H), the relationship between the calorific value of FC and the input fuel is expressed by the following linear expression.
From the heat storage unit, there is a decrease in heat due to natural heat dissipation, which can be assumed to be proportional to the amount of heat stored. Therefore, if the proportionality coefficient is η _FC (H), the amount of heat reduction due to natural heat dissipation is expressed by the following linear expression.
The heat storage amount Q (H, t) of the heat storage unit at the beginning of the period t is considered to have a lower limit and an upper limit, and the constraint conditions are set as follows.
The lower limit value Q _min (H, t) and the upper limit value Q _max (H, t) can reflect the setting value of the user for each time in addition to the case where it is determined by the specifications of the device.

Next, a constraint condition representing the transition of the heat storage amount of the heat storage unit is defined. At the beginning of the period t, the heat storage amount Q (H, t) of the heat storage unit is supplied by _FC q _FC ⁱⁿ (H, t), natural heat dissipation μ _FC (H, t), heat dissipation y ⁺ (H, t) by the radiator Further, due to the heat supply q _FC ^out (H, t) to the home, it changes to Q (H, t + 1) at the beginning of the next period t + 1. This is expressed by the following constraint equation.
Further, if δ _y (H, t) is a heat radiation by the radiator, 1 is 0, otherwise 0-1 variable, and if M is a sufficiently large number, variables y ⁺ (H, t) and δ _y ( The relationship of H, t) satisfies the following inequality.
Note that if the initial storage capacity of the heat storage unit is Q ₀ at the start of the optimization process span period, the following constraints are satisfied.

6.4.3.2.3.3. Objective Function FIG. 78 is a diagram illustrating a linear expression constituting the objective function. The objective function is defined as shown in FIG. 78 as the product of the following five linear expressions corresponding to costs and the risk amount calculated for each scenario i and the occurrence probability of the scenario, that is, the sum of expected values.

The objective function handled in this sub-module is given by the following equation as a weighted sum of cost and risk probability.
However, θ is a risk parameter set to a positive value. The degree of risk consideration can be controlled by controlling the magnitude of θ. For example, when θ is set to a large value, the minimization of the risk term is emphasized, and a risk-avoidance device control plan is calculated. For example, if the risk of running out of the battery is strongly considered, the control is to maintain a high power storage level. On the other hand, when θ is set to a small value, a control plan with an emphasis on costs such as purchase of grid power is calculated.

  FIG. 79 is a diagram illustrating an input parameter setting method for each optimization mode. The coefficients of each linear expression of the objective function are given as inputs of this submodule. At that time, according to the optimization mode designated by the user, appropriate values shown in the following table need to be set by the user or the administrator.

6.4.2.4. Creation of Optimal Plan FIG. 80 is a diagram illustrating the correspondence between output items and variables used in optimization. In this submodule, the three risk parameters θ _S (H), θ _L (H) and θ _U (H) designated as input values are set as the risk parameters θ of the objective function, and the optimization process is executed. By optimizing with three different risk parameters, three types of power storage plans are calculated for the battery. The relationship between output items and optimization variables is shown in the table below.

6.4.3. Control rule generation submodule 6.4.3.1. Concept 6.4.3.1.1. Overview of Function Devices to be controlled are EV storage battery, stationary storage battery (LIB), and fuel cell. The output of the optimization plan submodule contains these control plans. Based on these, a rule “control rule” for controlling each device is created. Based on the output of the optimization planning submodule, the basic state of each device in the planning mesh is defined, and the control rule of each control mesh is created based on the basic state. The basic state is a control state of the device in the planning mesh based on the optimization plan, and is invariable in the planning mesh when the optimization mode is other than the emergency mode. The optimization plan value of EV storage battery / stationary storage battery (hereinafter “storage battery” refers to both EV storage battery / stationary storage battery) includes storage amount management width (upper limit / lower limit of storage amount based on optimal plan) Therefore, the device is controlled so that the actual amount of power storage changes within this range as much as possible. The control rule is generated depending on the optimization mode. In particular, the control of the storage battery is a discharge control at a certain speed when connected to the system power supply, whereas in the emergency mode when disconnected from the system, the EV / PV cooperative power is controlled. Since the load following control is performed using the conditioner, the rule generation method is also greatly different.

  This system is based on the premise that either a stationary storage battery or EV is installed in one house, and the control rules generated by this submodule are also based on this assumption. If a plurality of storage batteries are installed, even if the management plan of each storage battery can be calculated by the optimization plan submodule, in order to configure a control rule accordingly, a hardware such as simultaneous control of multiple storage batteries is required. It is necessary to study a rule generation algorithm that also considers restrictions on wear.

6.4.3.1.2. Reason / Evidence The device control plan by the optimization plan sub-module minimizes the objective function set for each mode. Therefore, this module creates a control rule based on this result. In this module, while maintaining the device state calculated in consideration of the overall optimization by the optimization planning module, a control rule is constructed to maintain the power supply / demand balance at that time according to the optimization mode target.
For storage batteries, a management range for the amount of stored electricity is set. If charging / discharging is performed beyond this range, the situation will deviate greatly from the optimization plan. The storage battery is controlled so as not to deviate from the management width as much as possible.

The following restrictions exist in controlling a storage battery.
1) Charging / discharging of the storage battery is performed at a constant speed determined by the specification.
2) Excessive switching from charge to discharge and from discharge to charge will accelerate the deterioration of the rechargeable battery.
Consider manipulating devices in the planning mesh under these constraints.

6.4.3.2. Input / Output FIG. 81 is a diagram illustrating input / output of the control rule submodule. As shown in FIG. 81, the submodule input is an EV storage battery storage plan, a stationary storage battery storage plan, and a fuel cell power generation plan in each plan mesh. These values are part of the output of the optimization planning submodule. Using these values, control rules for EV storage batteries, stationary storage batteries, and fuel cells for each control mesh are created and used as output.

6.4.3.3. Creation of rules for system connection The control rules are created by obtaining the basic state of devices in each planning mesh from the optimization plan.
6.4.3.1.3.1. Creation of Basic State The device state in each planned mesh derived by the optimization plan is called a basic state.

  FIG. 82 is a diagram illustrating a basic state of the device. In the storage battery, discharging is performed when the discharge power amount is positive in the planned mesh, charging is performed when the charging power amount is positive, and standby is performed when both values are 0 (meaning that neither charging nor discharging is performed). The fuel cell is activated when the fuel cell power generation amount is positive, and stopped when it is zero.

6.4.3.2.3.2. Control rule creation from EV storage battery charge / discharge plan From the EV storage battery charge / discharge plan and the basic state of each planned mesh, a control rule for the EV storage battery at time t in the planned mesh is created.

  First, the basic state in each planned mesh is obtained from the EV storage battery discharge amount and the charging power amount as shown in FIG. In the same plan mesh, in order to prevent excessive charging / discharging switching of the storage battery, it is not discharged when the basic state is charging, and conversely, it is not charged when the basic state is discharging. The EV storage battery charge / discharge plan includes the storage amount of the EV storage battery and the management range indicating the upper limit and the lower limit thereof, and it is not desirable that the measured value of the EV storage power amount exceed this range. Therefore, control is performed to keep the actual value of the amount of electricity stored in the EV storage battery within this range as much as possible.

  FIG. 83 shows an example of control when the basic state is discharging. FIG. 84 is an example of control when the basic state is charging. The EV storage battery is put into a discharged state by the first control rule of the planned mesh (12:00 control rule). Thereafter, when the measured value of the charged amount falls below the lower limit of the management width, the EV storage battery is set in a standby state. Next, when the measured value of the charged amount exceeds the upper limit of the management width, the EV storage battery is discharged again. When the basic state is discharging, only discharging / standby is allowed, and charging is not performed. Similarly, when the basic state is charging, only discharging / standby is allowed and discharging is not performed (FIG. 84). In order to realize this, a control rule is created based on the following processing.

6.4.2.3.2.1. Calculation of management width in the planning mesh The management width in each planning mesh is calculated. The value of the management width obtained by the optimization plan is that at the planning mesh start time, and the management width is not defined at time t in the planning mesh. Therefore, the management width at time t is calculated based on the management width of the planned mesh and the management width of the next planned mesh. In the planning mesh, it is assumed that the upper limit of the management width _changes at a certain rate, and a straight line connecting the upper limit value V _1max of the planning mesh and the upper limit value V _2max of the next planning mesh is set as the upper limit at time t. . Similarly, the lower limit of the control range, the straight line connecting the lower limit V _2min lower limit V _1min and next plan mesh of the plan mesh and the lower limit value in the control mesh.

The upper limit value / lower limit values v _max (t) and v _min (t) at the time within each planned mesh, where the planned mesh width is 1 hour, can be obtained by the following equations.

6.4.3.3.2.2. Control rule creation method At each point in the mesh, a control rule is created according to the following rules.
(A) When the basic state is discharging (a) When the measured value of the EV storage amount is within the management range or exceeds the upper limit of the management range, the EV storage battery is discharged (b) When the measured value falls below the lower limit of the management width, the EV storage battery is in a standby state. (A) When the basic state is charging (a) The measured value of the EV storage amount is within the management width, or the lower limit of the management width (B) When the measured value of the EV storage amount exceeds the upper limit of the management range, the EV storage battery enters a standby state.
(C) When the basic state is standby (a) Regardless of the management width, the EV storage battery is in the standby state

FIG. 85 is a diagram illustrating EV storage battery storage plan values and basic states.
FIG. 86 is a diagram illustrating a storage battery control rule created from FIG. 85.

6.4.3.3.2.3. Creation of control rules for stationary storage batteries Create control rules for stationary storage batteries in the control mesh from the stationary storage battery charge / discharge plan and the basic state of each planned mesh. A method for creating a control rule for a stationary storage battery follows that of an EV storage battery.

6.4.3.3.2.4. Creation of Control Rule from Fuel Cell Charging / Discharging Plan FIG. 87 is a diagram illustrating fuel cell control for each optimization mode. If the basic state of the fuel cell is in operation, its control is determined depending on the optimization mode. In the economic mode, it is necessary to control so that the power cost is as low as possible. Therefore, compare the power generation unit price of the grid and the fuel cell for each planned mesh, and if the power generation unit price of the fuel cell is low, it is better to use fuel cell power generation with priority. Do not use control. Conversely, if the grid power unit price is lower, the output of the fuel cell is limited to the lowest level, and control is performed so that the shortage of power is covered by the grid. In the self-sufficiency mode, it is required to perform control that uses as little grid power as possible. Therefore, the fuel power is controlled by the load following mode to minimize the system power. In the environmental mode, it is necessary to reduce the carbon dioxide emission as much as possible. Therefore, the power generation unit price in the economic mode is changed to the carbon dioxide emission, and the priority between the fuel cell and the system power is determined. In other modes, when both fuel cell and grid power are available, there is no difference in priority between power generation methods, so either economic mode or self-sufficiency mode is registered as the default mode. .

6.4.3.4. Creation of rules in emergency mode When grid power is disconnected and emergency mode is selected, a storage battery with a large charge / discharge capacity plays a role of keeping the voltage in the house constant on behalf of the grid. For this purpose, operation is performed in a mode in which the load is followed according to the power demand in the house, instead of controlling the storage battery by designating a constant charge / discharge capacity. In the load following mode, if the in-house power becomes surplus, the surplus is charged, and if it is insufficient, the deficient discharge is performed. Therefore, in the emergency mode, the device to be controlled is only the fuel cell, and it is necessary to maintain the power storage level within the management range calculated by the optimization plan submodule by controlling the power generation amount.

6.4.4.1.4.1. Generation of Fuel Cell Control Rules FIG. 88 is a diagram illustrating control of the fuel cell when the storage battery is in a discharged state. FIG. 89 is a diagram illustrating control of the fuel cell when the storage battery is in a charged state. As shown in these figures, a control rule for the fuel cell is created according to the following rules at each time in the planned mesh.
(A) When the measured value of the EV or stationary LIB charged amount is below the lower limit of the management width, the output of the fuel cell is increased.
(A) When the measured value of the EV or stationary LIB charged amount exceeds the upper limit of the management width, the output of the fuel cell is decreased. However, the output is not less than the minimum output (that is, stopped state).
(C) When the measured value of the EV or stationary LIB charged amount is within the management range, the output of the fuel cell is not changed (the previous output is maintained).

7). Outline From Chapter 7 onwards, the items explained in Section 6 will be explained in more detail.

7.1. Configuration of control algorithm 7.1.1. Positioning of Control Algorithm from the Perspective of Peripheral System FIG. 90 is a diagram illustrating the cooperation between this system and its peripheral system. This system operates in the information management server, receives it via a rule update request optimization engine I / F from an external system, and executes control rule generation processing. The control rule generation process consists of a main part and an algorithm part, predicts the future state of each device using past performance data in the analysis DB, and minimizes CO2 emissions under given constraints The optimization calculation is performed based on an index such as “Yes”, and a rule for optimally controlling the device in the HEMS house is created. The created rule is stored in the analysis DB.

  Refer to Section 10 (External Sequence) for the trigger of optimization processing and rule creation processing of this system. A rule update request from the outside of this system is made via the optimization engine I / F. Refer to section 11 (External IF) for calling the rule update request between the optimization engine and the external module.

  After the optimization process / rule creation is completed, the optimization engine I / F notifies the optimization engine I / F that the rule creation process has been completed, and the optimization engine I / F receives the notification to the home server in each HEMS house. Notice. The home server receives the notification and obtains the latest rule. This system executes a rule creation process triggered by a rule creation process request from an external module, and the rule is updated when necessary in the HEMS house by detecting the end event of the rule creation process on the request sending side. Device control using the latest and optimum control rules is possible. Since it is expected that it will take some time to complete the control rule creation process, the control rule creation process is executed asynchronously with the rule update request. Further, the optimization engine I / F mediates a request to the optimization engine, thereby ensuring portability at the time of transition from development STEP1 to development STEP2. In addition, expandability can be secured in preparation for the enhancement of the optimized engine server.

7.1.2. Function Division FIG. 91 is a diagram illustrating a conceptual function division configuration of a control algorithm. FIG. 92 is a diagram illustrating functional division of the control algorithm. First, this system is divided into a main part, an algorithm part, and an analysis DB part. The main part is composed of main function classifications, the algorithm part is composed of four function classifications of demand, supply, power storage, and optimum supply and demand, and the analysis DB section is composed of analysis DB function classifications.

  The demand, supply, and power storage function classifications virtually reproduce the crisis configuration of HEMS, and define the relationship between the demand amount, supply amount, and power storage amount of each device and other observed values. These functional categories define preconditions for the operation of each device to be managed.

  In the optimum supply and demand plan function category, an operation plan for the equipment that maximizes or minimizes the objective function corresponding to the optimization mode is created under the restrictions defined in the demand / supply / capacitor function category. In addition, a control rule for each device for realizing the calculated operation plan is created.

  The main function division realizes integration of these function divisions and linkage with an external system. The analysis DB holds observation values related to each device, other data necessary for optimal control, intermediate data being processed, control rules serving as final outputs, and the like. Although the construction of the analysis DB is out of the scope of design / implementation, the logical design of the table used by this system on the analysis DB is targeted for design / implementation.

8). Overall Configuration FIG. 93 is a diagram exemplifying subdivided function divisions. In this chapter, the functional categories shown in Section 7 (Overview) are broken down into smaller logical functions, and the roles played by each logical function are clarified. The main function classification can be divided into a main control for processing and a control DB for storing information related to the control. Demand / supply / storage function divisions are broken down into devices or sets of devices. The demand function category is divided into home appliance power demand, EV power demand, and heat demand, the supply function category is divided into PV supply, system power flow, fuel consumption, hot water storage, and the storage function category is divided into EV storage and storage battery storage logical functions. . The optimized supply and demand plan function classification can be divided into an optimization plan for calculating optimal control and a generation of control rules for determining how and at what time each device is controlled based on the optimal plan.

  The analysis DB stores data used for analysis or control information for performing analysis processing. The data to be analyzed are the supply / demand actual value of each device in the house and the actual value of the travel data, and a life log DB and a travel data DB for storing these are created. There are meteorological information and a calendar as factors affecting the change of the actual value to be analyzed, and a weather DB / calendar DB is created in order to store past / future weather information / calendar. Control rules for each device are created as the final output of this system. A rule DB for storing this is required. The optimization setting DB stores parameters serving as input for optimization processing. The statistics / optimization DB stores intermediate data generated in the process of creating a control rule. The analysis DB unit also includes a control DB used for main control.

9. Method 9.1. Outline of method In this chapter, the implementation method of the entire system is determined on the assumption that this system is implemented, and the functions of each logical function described in section 3.2 (overall configuration) are specified. The embodied logical functions are called physical functions, and they are the targets of implementation. A logical function may be divided into multiple physical functions. Some logical functions do not need to be implemented or are integrated into other physical functions. No physical function is created for these logical functions.

  FIG. 94 is a diagram illustrating physical functions of the present system and their functions / roles. FIG. 94 shows physical functions and their functions / roles in this system based on the determination of the mounting method in this chapter.

  FIG. 95 is a diagram illustrating a system configuration diagram. The linkage between this system and the optimization engine IF will be described in Section 10 (External Sequence). The external IF is described in section 11 (External IF). Details of physical functions (functions) excluding the external IF in the figure are described in section 13 (functions), and details of each DB are described in section 12 (data flow).

9.2. Execution of external calls and processes in each physical function of the main part The main part is composed of a main process and an external IF, and uses a control DB. This system accepts a control rule creation request by an HTTP request from the optimization engine IF. The state of information / control rule creation processing obtained by the control rule creation request is stored in the control DB. In the main process, each process of the algorithm part is called, and finally a control rule for each device is generated. After the processing of the control rule is finished, the optimization engine IF is notified that the processing is finished.

9.3. Demand / supply prediction and optimization in each physical function of the algorithm part 9.3.1. Demand / supply forecasts for each physical function in demand function category and supply function category Electricity demand in the home, EV usage, hot water (heat) consumption, PV power supply, etc. Changes due to various factors such as lifestyle. For this reason, it is difficult to predict the future amount of electric power consumption of home appliances, but it is considered to improve the prediction accuracy of the future amount of demand by using past actual values. Predicted values are obtained mainly by statistical methods. If this is difficult, a value equivalent to the predicted value is calculated by another method. When calculating a predicted value by a statistical method, the development efficiency can be improved by using statistical software including various statistical libraries as compared to the development using a normal programming language.

9.3.2. Optimization in each physical function of the optimum supply and demand planning function division 9.3.2.1. Optimization plan creation Optimization using the stochastic mixed programming method in the physical function The optimization plan is based on the future transition (predicted value) of household appliance power demand, EV power consumption, heat demand, and PV supply, and demand. -Create a control plan for each device, taking into account that the forecast of supply volume will be somewhat out of order. This can be realized by the stochastic mixture programming method, and this method is used in this system. The optimal solver is used when the control plan is obtained by the stochastic mixed programming method.

  By using probabilistic mixed programming, it is possible not to estimate the predicted value of each demand / supply quantity at a certain point in the future as a point, but to take into account how much the predicted value varies, even when it varies Create a control plan for equipment that satisfies various operational restrictions as much as possible. Such a “predicted value with variation” is expressed by a combination of one or more scenarios (demand scenarios or supply scenarios) and their occurrence probabilities, and these are input to the optimization plan. The (demand or supply) scenario represents how the predicted value changes along the time axis. The probability of occurrence indicates the degree of accuracy with which the predicted value represented by the scenario can change.

9.3.2.2. Necessity of physical function to create an integrated scenario Since the optimization plan function takes the supply / demand scenario and its occurrence probability as input, the output of household electrical power demand, EV power demand, heat demand, and PV supply is also a demand / supply scenario. This is the probability of occurrence. The scenario assumed by the optimization planning function must take into account concurrency. However, it is assumed that the scenarios output by the functions described above and their occurrence probabilities occur as independent events. Therefore, an integrated scenario creation function is introduced. In the integrated scenario creation function, scenarios and occurrence probabilities on the assumption of non-simultaneous occurrence are input, and combinations of scenarios that consider simultaneous occurrence called integrated scenarios and their occurrence probabilities are output.

9.4. Integration of functions For some logical functions, the relationship between variables is clearly specified as a mathematical expression. Since these relational expressions are used as they are as constraints of the optimization plan, their implementation is integrated into the optimization plan physical function. This corresponds to the logical functions of system power flow, fuel consumption, hot water storage, EV storage, and storage battery storage.

10. External sequence 10.1. Start of control rule generation process 10.1.1. Activation of Control Rule Creation Process FIG. 96 is a diagram illustrating a sequence diagram related to the control rule generation process. In response to an HTTP request from the optimization engine I / F, a control rule generation process is started, and a process start notification is returned to the optimization engine I / F as a process reception response. After that, at the timing when the process is completed, a process completion notification is sent to the optimization engine I / F as a request by HTTP. It is assumed that the optimization engine is called about four times a day for one house.

10.2. Processing Trigger FIG. 97 is a diagram illustrating events that can be processing triggers. The processing trigger of this system can be the following cases.
A) Periodic update of control rules B) When input values have changed since the start of the last rule generation process For these events, the degree of influence of the events (whether the preconditions for generating control rules are greatly destroyed) -Considering the occurrence frequency and detectability, it is determined whether or not to be a processing trigger. The degree of influence on a rule indicates how much the rule changes when a control rule is updated from an event. An event having a large degree of influence is preferably a processing trigger. Since the control rule generation process takes a certain amount of processing time, if the update frequency is high, the system is loaded. Therefore, an event with a high occurrence frequency is not a trigger for processing. An event with a medium update frequency is used as a processing trigger, but if the frequency can be controlled, it should be low. Events that cannot be detected are not triggered. For events that belong to "when user settings are changed" or "when administrator settings are changed", it is necessary to determine whether to trigger the process after considering the impact on life. is there. Even if the update of the rule due to the event “change in optimization mode” or “change in optimum storage amount” is not immediately reflected, it does not hinder the user's life. On the other hand, “addition / change of EV reservation” and “addition / change of full hot water reservation” impede life, such as being unable to use EV or not taking a bath. The former is not an opportunity, and the latter is an opportunity. However, regarding “addition / change of EV reservation”, there is a possibility that triggers frequently occur depending on the EV usage form. Therefore, if multiple triggers occur within a certain interval, processing such as ignoring except one is necessary. It becomes.

11. External IF
11.1. Control rule generation process start argument and return value 11.1.1. Control rule generation process activation argument and return value 11.1.1.1. Process Activation Request FIG. 98 is a diagram exemplifying control rule generation process activation arguments. Acquired as a GET parameter of the HTTP request.

11.1.1.2. Acceptance Notification Response FIG. 99 is a diagram illustrating an acceptance notification state. The acceptance status is output as text data of the HTTP response. If the argument is incomplete, the return value is returned as reception failure. Details of error check are described in 13.1.3. Refer to error check specifications.

11.1.1.3. Processing Completion Notification Request FIG. 100 is a diagram illustrating an acceptance notification state. The processing completion status is output as a GET parameter of the HTTP request.

11.1.1.4. Processing Completion Notification Acceptance Response FIG. 101 is a diagram illustrating an acceptance notification state. The reception state is acquired as text data of the HTTP response.

12 Data Flow FIG. 102 is a diagram illustrating a data flow in this system. The various DBs are divided into logical tables, and the input / output relationship between the data including those tables and each function is clarified. In the control algorithm call request, the house ID / start date / time is specified as an argument. The main process receives these pieces of information and temporarily stores them in the control DB. The control DB includes a house master and control rule creation process management. The main process acquires information necessary for execution from the control rule creation process management table, and passes the control rule creation process ID, the control rule creation period start date and time, and the control rule creation period end date and time as arguments when each function is called. It is called by specifying the start date and time of the control rule creation process / house ID.

  In the weather DB, there are tables of weather information results, hourly weather information forecasts, and daily weather information forecasts. Alternatively, the life log DB includes tables for household electrical power demand results, heat demand results, and PV supply results. In the travel data DV, there are EV power demand results and travel reservations. There is a calendar table in the calendar DB. These tables are input to the power demand scenario creation function, the heat demand scenario creation function, the PV supply scenario creation function, and the EV power demand scenario creation function. The simulated results table in the weather DB, life log DB, and travel data DB stores artificially created results, and each scenario creation function even in situations where there is no or insufficient results for each device In order to guarantee operation or a certain degree of prediction accuracy. These four scenario creation functions output to the scenario master / scenario table.

  The statistics / optimization DB includes a scenario creation process, a scenario relation table serving as an input / output of the optimization process, and an optimization plan relation table. The scenario relation table includes a scenario master, a scenario, and an integrated scenario master. The optimization plan related tables include an EV storage battery plan, a stationary storage battery plan, a grid power amount plan, a heat storage unit heat amount plan, a heat plan, an EV power amount plan, a LIB power amount plan, and an FC heat generation amount plan table.

  The integrated scenario creation function takes a scenario master and a scenario as input, and outputs to each table of the integrated scenario master and integrated scenario. The optimized plan creation function takes the integrated scenario master, integrated scenario, and tables in the optimization setting DB (optimization parameters, time series optimization parameters, optimization engine parameters) as input, and outputs to the optimization plan related tables Do. The control rule DB is composed of a control rule table. The control rule creation function receives the EV power amount plan, the LIB power amount table, and the FC heat generation amount table as inputs and outputs them to the control rule table.

Depending on the scale of operation, the system may store more than tens of millions of records (see section 19.4 for an estimate of the number of records stored in each table). The following three measures are taken to avoid performance degradation due to an increase in the number of records.
A) Partitioning Perform partitioning for each table. The number of records stored in one table is prevented by changing the table to be stored according to the column value according to the date.
B) Saving intermediate data The table belonging to the statistics / optimization DB is for storing intermediate data generated during the processing of this system. These data are not used for purposes other than verification after the processing is completed. For each of these tables, a backup table is prepared, and the data generated in the process is transferred to the backup table immediately before the process ends. Tables for saving intermediate data are shown in Section 19.3.
C) Saving of actual data Actual data is assumed to refer to the past two years, and data before that is not used in each process. Therefore, by periodically backing up and deleting unnecessary data, the performance degradation of the performance table is prevented. Tables for saving actual data are shown in Section 17.3.

13. Function 13.1. Main processing 13.1.1. Functional Overview FIG. 103 is a diagram illustrating a main process IO related diagram. The main process receives an external request, and manages and executes a control rule creation process corresponding to the request. See section 11 (External IF) for external call IFs that trigger main process calls. In response to the process activation request (HTTP), the main process issues a control rule creation process ID and returns an acceptance notification response (HTTP) to the request. In addition, each prediction / optimization function is executed in asynchronous processing, and after completion of processing, a processing completion notification request (HTTP) is transmitted to the optimization engine I / F, a processing completion notification reception response is received, and a control rule Complete the creation process.

  FIG. 104 is a diagram illustrating a processing order flow diagram. As shown in FIG. 104, the control rule creation process executed by the main process calls each function in the demand module / supply module / optimum supply / demand planning module.

13.1.2. Arguments and return values The following describes the arguments (process activation request parameters) and return values (acceptance notification response parameters) of the main processing function. Further, the processing completion notification request and the processing completion notification reception response that are notified to the optimization engine I / F after the processing is completed will be described below. .

13.1.2.2.1. Argument 11.1.1.1. See process start request.

13.1.2.2.2. Return value 11.1.1.2. See reception notification response.

13.1.2.2.3. Processing Completion Notification Request 11.1.1.3. See processing completion notification request.

13.1.2.4. Processing completion notification acceptance response 11.1.1.4. Refer to processing completion notification acceptance response.

13.1.3. Error Check Specification FIG. 105 is a diagram illustrating an error check list. FIG. 105 shows an error check specification of the main processing function. If it corresponds to the error check, the requested processing is interrupted and the acceptance notification response is returned to the request source as acceptance failure. The details are 11.1.1.2. See reception notification response.

13.1.4. DB editing specification The following shows the acquisition and storage of values from the database in the main processing function.
13.1.4.1. Acquired data list The acquisition of values to the database is shown below.
FIG. 106 is a diagram illustrating a list of acquired data.

13.1.4.2. Stored Data List FIG. 107 is a diagram illustrating a stored data list.

13.1.4.3. Update Data List FIG. 108 is a diagram illustrating an update data list.

13.1.5. Processing details 13.1.5.1. Processing Flow FIG. 109 is a diagram illustrating a flow of main processing. In the main process, a control rule creation process ID is issued in response to a process activation request (HTTP), and an acceptance notification response (HTTP) to the request is returned according to the following flow. In addition, each prediction / optimization function is executed in asynchronous processing, and after completion of processing, a processing completion notification request (HTTP) is transmitted to the optimization engine I / F, a processing completion notification reception response is received, and a control rule Complete the creation process. Save them in the database.

13.1.5.2. Confirmation of Processing Status FIG. 110 is a diagram exemplifying processing status update contents. With reference to the control rule creation process management table using the house ID and start date / time received as parameters as acquisition conditions, if the process status of the record with the newest process date / time is 2 (executing), the process status of the same record is 3 (suspended) Update to middle). If there is no record, nothing is updated.

13.1.5.3. Storage of Control Rule Creation Process Data FIG. 111 is a diagram exemplifying process status update contents. The house ID and start date / time received as parameters are stored in the control rule creation process management table. At the same time, the system date and time acquired by the application at the processing date and time, 2 (processing) in the processing status, and the current date as the creation date and time are stored. Note that the system date and time stored in the processing date and time is retained for use in later processing.

13.1.5.4. Acquisition of the assigned control rule creation process ID The control rule creation process ID is obtained from the control rule creation process management table. The acquisition conditions are the house ID and start point date and time received as parameters as the house ID and start point date and time in the control rule creation process management table, respectively, and the system date and time acquired in the storage process of the control rule creation process data as the process date and time. .

13.1.5.5. Reply of the reception notification response The control rule creation process ID acquired in the pre-processing is edited and returned as the reception notification response in the HTTP response body.

13.1.5.5.6. Check and control the number of process executions Refer to the control rule creation management table, compare the number of records with process status 2 (running) and the maximum number of simultaneous executions of control rule creation from property information, and the number of records is the maximum number of simultaneous control rule creation If the number of executions is greater than or equal to the number of executions, the subsequent process execution is waited. In addition, when the number of records becomes less than the maximum number of simultaneous execution of control rule creation, control is performed so that the waiting control rule creation processing is sequentially started in the order of requests.

13.1.5.5.7. Acquisition of property information Processes after this process are executed asynchronously with the response of the reception notification response. Get the property information defined in the system property file and create an object to hold each as a constant. See section 19.1 for property information.

13.1.5.8. Determination of control rule creation period start point date and time and control rule creation period end point date and time From the acquired start point date and time, the control rule creation period start point date and time and the control rule creation period end point date and time are determined. As for the start date and time, the latest target date and time is obtained for each table of household electric power demand record, EV power demand record, heat demand record, PV supply record, and weather information record, using the house ID as an extraction condition. Of these, for the oldest target date and time, the date and time after rounding off the minutes is set as the control rule creation period start date and time. Further, based on the start date and time of the parameter, the control rule creation period end date and time is set to 24:00 of the date added with the span length acquired from the property information.

13.1.5.9. Creation of Input Data FIGS. 112 to 115 are diagrams illustrating input data reference sources and storage destinations. Prior to the execution of the control rule creation process, user settings, administrator settings, fuel / system power supply price and other values are converted into a format suitable for scenario creation / optimization process. In addition, it is necessary to clarify the location and details of the reference sources marked with * in the table. These data must be up-to-date before the control rule creation request is issued.

13.1.5.10. Execution of Each Function FIG. 116 is a diagram illustrating a processing order flowchart (repost). Each prediction / optimization / control rule creation function is sequentially executed using the control rule creation process ID, the control rule creation period start point date and time, and the control rule creation period end point date and time as parameters. After executing each prediction / optimization / control rule creation function, the execution result is checked each time, and if it ends abnormally, the subsequent functions are not executed. Also, if the process has not been completed when the prediction optimization process timeout period has elapsed since the start of process execution, it is treated as a timeout. At that time, the subsequent prediction / optimization / control rule creation function is not executed. Note that the processing status of the control rule creation management table is referred to during the function execution, and the prediction / optimization / control rule creation function after the status 3 is not executed.

13.1.5.11. Data evacuation Records are compared against tables that are concerned about the number of records in the table. For the table to be saved, the records stored with the request as the trigger (same control rule creation process ID) are extracted and deleted, and stored in the corresponding backup table. The table to which data is saved is a table having the control rule creation process ID as a primary key, and the number of records exceeds 10 million records when used continuously (see FIG. 212 for the estimated number of tables for each table). . The tables to be obtained are: scenario master, scenario, integrated scenario master, integrated scenario, EV storage plan, stationary storage battery storage plan, EV energy plan, stationary storage battery storage plan, system power plan, heat storage unit heat plan, FC The power generation plan, the heat plan, the time series EV travel reservation, the optimization engine parameters, the time series optimization parameters, and the optimization parameters.

13.1.5.12. Updating Process Status FIG. 117 is a diagram illustrating a process status category list. FIG. 118 is a diagram exemplifying processing status update contents. After executing the function, the process status of the control rule creation process ID is updated. Update the process status according to the process completion result.

13.1.5.13. Send processing completion notification request Send processing completion notification request after execution of the function is completed. A processing completion notification request URL is acquired from the property information, the processing result is edited into the GET parameter of the HTTP request for the URL, and the request is transmitted. For the GET parameter of the processing result, refer to section 14 (external IF processing completion notification acceptance response).

13.1.5.14. Receive processing completion notification acceptance response Receives response to processing completion notification request. At that time, if the HTTP request times out, a processing completion notification request is transmitted up to three times, and if both times out, an error is assumed.

13.2. Demand forecast function 13.2.1. Electric appliance demand scenario creation 13.2.1.1.1. Functional Overview FIG. 119 is a diagram illustrating a home appliance power demand scenario creation IO relation diagram. Generation of household electrical power demand scenario, forecast error, percentile value, household electrical power demand scenario in specified housing / period based on household electrical demand, temperature / weather, season, calendar information, weather information, weather information forecast Get probabilities and store them in the database.

13.2.1.2. Arguments and return value 13.2.1.2.1. Argument FIG. 120 is a diagram illustrating an argument of the home appliance power demand scenario creation function.

13.2.1.2.2. Return Value FIG. 121 is a diagram illustrating a return value of the home appliance power demand scenario creation function.

13.2.1.3. Error Check Specification FIG. 122 is a diagram illustrating an error check specification. FIG. 122 shows an error check specification of the home appliance power demand scenario creation function. When it corresponds to the error check, the process is interrupted and a return value is returned to the main process as a process failure. For details, see 13.2.1.2.2. See return value.

13.2.1.4. DB Editing Specification Acquisition and storage of values from the database in the home appliance power scenario creation function will be described.
13.2.1.4.1. Acquired Data List FIG. 123 is a diagram illustrating acquisition of values from the database.

13.2.1.4.2. List of Stored Data FIG. 124 is a diagram illustrating storage of values in the database.

13.2.1.4.3. Update data list This function does not update the database.

13.2.1.5. Processing details 13.2.1.5.1. Processing Flow FIG. 125 is a diagram for explaining a home appliance power demand scenario creation processing flow. In the home appliance power demand scenario creation process, the occurrence probability of home appliance power demand scenario / prediction error / home appliance power demand scenario is obtained by the following flow and stored in the database.

13.2.1.5.2. Definition of season The season consists of three parts: winter, summer, spring and autumn. Each period of winter, summer, spring and autumn is obtained from parameter information.
13.2.1.5.3. Definition of date category Implement the function to create date category. The date category is defined as follows.
For holidays (2 categories): Presence of precipitation (2 categories)
If it is not a holiday (14 categories): Precipitation (2 categories) x Day of the week (7 categories)

Presence or absence of precipitation is determined as follows: A) When the target date belongs to the actual period The precipitation is obtained from the meteorological information actual table. If the precipitation is greater than 0, there is precipitation. If the precipitation is 0, there is no precipitation. And
B) When the target day belongs to the forecast period, and the hourly weather information forecast table has precipitation probabilities for each time zone on that day, obtain the precipitation probability for each time zone on that day, and the time zone when the precipitation probability is greater than or equal to the threshold If there is at least one, it is raining, otherwise it is not raining. The threshold is acquired from the parameter information.
C) When the target date belongs to the forecast period and the daily weather information forecast table has a precipitation probability for the day. The threshold value is the same as the value used in B above.

  The days of the week are 7 categories of Sunday, Monday, Wednesday, Wednesday, Thursday and Friday. Whether or not the target day is a holiday is determined based on whether or not a record corresponding to the day exists in the calendar table.

13.2.1.5.4. Determination of Forecast Period / Actual Period The forecast period means a period in which the electric home appliance power demand value for each plan mesh is targeted for prediction. The prediction period is from the control rule creation period start date and time given as an argument to the control rule creation period end date and time (including the start point and not the end point). The actual period is a period having data used for learning the prediction model. The actual period is a set of days having household electric power demand actual results in all the planned meshes in the household electric power demand actual result table (the days do not have to be consecutive). However, data from past 2 years will not be used.

13.2.1.5.5. Creation of objective variable for learning By the following processing, a deviation value from the average electric home appliance power demand which is the objective variable of the learning data is calculated. For all planning meshes during the forecast period: A) The day to which the plan mesh belongs is assumed as the forecast date. Let P be the date category of the forecast date.
B) The actual value of the household electric power demand amount for the day belonging to the same date category P as the forecast date among the dates of the actual period is acquired. The acquisition process of the actual value of the household electric power demand is in accordance with the process of 13.2.1.5.6.
C) The average value of the same planned mesh time zone value is obtained for the actual value of the acquired household electric power demand. This is called the average power demand aveE _H (H, P, t). Here, H represents a house, and t represents a time zone of the planned mesh.
D) The household electrical power demand E _H (H, P, S, t) for the dates belonging to the same category P and the same season S as the forecast date in the actual period is acquired.
E) About the value calculated | required by D, the difference y (H, P, S, t) of the household appliance power demand of the same plan mesh as the said plan mesh and a category average power demand is calculated | required. This is called the home appliance demand divergence value, and is obtained by the following equation.
According to the above processing, home appliance demand amount deviation values for each plan mesh are obtained by the number of category actual days.

13.2.1.5.6. Acquisition of Home Appliance Power Demand Results with Planned Mesh Width The actual value of home appliance power demand is accumulated as an aggregate value every 10 minutes. The planning mesh is variable and is obtained from parameter information. In the processing within this function, the actual value of the household electric power demand amount in units of planned mesh is required. Therefore, when the total width of the actual value is different from the planned mesh width, the home electric power demand actual value is obtained by recalculating the home electric power demand actual value in the home electric power demand actual table with the total width of the plan module. Here, it is assumed that the planned mesh width is longer than or equal to the total width of the actual values.

13.2.1.5.7. Creation of explanatory variables for learning By the following processing, a deviation value from the actual temperature and its average, which is an explanatory variable of the learning data, is calculated. For the planning mesh during the forecast period,
A) The day to which the planned mesh belongs is set as the predicted date.
B) Acquire the actual temperature m (P, S, t) of the day belonging to the same category P and the same season S as the predicted date. The temperature is obtained from the weather information performance table.
C) The planned mesh average temperature avem (P, S, t) of the planned mesh is obtained from the actual temperature m (P, S, t).
D) Difference between the actual temperature and the average temperature of the planned mesh according to the following formula
Ask for. This is called the temperature record divergence value. By the above process, the temperature divergence value for each plan mesh is obtained by the number of actual values acquired by the process of i.

13.2.1.5.8. Creation of explanatory variables for prediction (calculation of deviation value of predicted temperature)
The divergence value between the predicted temperature and the average value is calculated by the following processing. For all planning meshes during the forecast period,
A) The day to which the planned mesh belongs is set as the predicted date.
B) The predicted temperature value m ′ (t) in the planned mesh is acquired by the process of 13.2.1.5.9.
C) Difference between temperature and average temperature by the following formula
Ask for. This is called the temperature forecast divergence value. By the above processing, the temperature forecast divergence value for each plan mesh is obtained by the number of category actual values.

13.2.1.5.9. Obtaining predicted temperature change Obtain the predicted temperature change by the following process.
A) When there is hourly weather forecast information The latest predicted temperature for every hour (= for each planned mesh) is acquired from the hourly weather information forecast table.
B) When there is only weather forecast information by date For the specified forecast date,
i. Get the latest forecast maximum temperature and forecast minimum temperature on the forecast date. From the daily weather information forecast table, the highest temperature and the lowest temperature on the forecast day are acquired and set as the predicted highest temperature and the lowest forecast temperature.
ii. From the performance period, obtain the highest and lowest temperature results for all days and calculate the following values.
iii. N days are selected in ascending order of the above values, the temperature results for that day are obtained from the weather information results table, and the average value is calculated for each planned mesh. This value is used as the predicted temperature transition for the day. The value of N is obtained from parameter information.

13.2.1.5.10. Predicted value creation (in summer and winter)
When the forecast date belongs to the summer / winter category, the forecast value of the power demand in each plan mesh is obtained as follows. The previously created demand divergence value Y _i is the objective variable, and the actual temperature divergence value X _i is the explanatory variable.
Assuming this relationship, simple regression is executed to estimate the intercept α ^ (S) and the explanatory variable coefficient β ^ (S). Using these values, an estimated value y ′ (t) of the home appliance power demand amount divergence value in each plan mesh on the forecast date is obtained from the temperature forecast divergence value X ′ (t).
The average demand amount aveE _H (H, P, t) is added to this to obtain a predicted value of the power demand amount in each plan mesh.

13.2.1.5.11. Predicted value creation (for spring and autumn)
A) Preparation FIG. 126 is a diagram illustrating the relationship between the demand for home electric power and the temperature in spring and autumn. As shown in FIG. 126, it is assumed that the sensitivity of the objective variable changes depending on the temperature level. _{Incidentally, y m (H, P,} S, t) denotes power demand y Condition 2 (within an average ± T degree) (H, P, S, t) the average.

In other words, each state in the figure is
It is expressed. This is in average temperature, within the range deviation of ± T ° C. from the basic demand patterns, sensitivity to temperature change is little, the average y _m of power demand in the state 2 (H, P, S, t) in It means that it is assumed to be constant.

  FIG. 127 is a diagram illustrating another example of the relationship between the electric home appliance power demand and the temperature in spring and autumn. Now, assuming that the sensitivities to the temperature changes in state 1 and state 3 are equal, if the sign of the power demand (deviation from the average) in state 3 is reversed, FIG. 126 is transformed as shown in FIG. And the relational expression of electric power demand and temperature is estimated by estimating temperature sensitivity from FIG.

The inclination of the straight line at the center in the figure is β. Under the above-mentioned assumption, as shown in FIG. 127, when the straight line in state 3 in state 1 is shifted vertically by α1 = βT and α2 = −βT, respectively, it coincides with the straight line in the center portion. To estimate the parameters α1, α2, and β, the following log likelihood lnL may be maximized.
A numerical value i (i = 1, 2, 3) below the Σ symbol represents a state. This means that discrete data is divided into three sections and summed.

Differentiating this likelihood function with the estimated parameter and setting it to zero, the following relational expression is obtained.

B) Parameter Estimation Parameters α1, α2, and β are obtained by the following procedure.
i. Appropriate initial values α1 ^(k) , α2 ^(k) , β ^(k) , and σ ^(k) (k = 1) are given to α1, α2, β, and σ.
ii. α1 ^(k) , α2 ^(k) , β ^(k) , σ ^(k is substituted into likelihood, and log likelihood lnL (k) is calculated.
iii. Parameters α ^ 1, α ^ 2, β ^, and σ2 ^ are obtained from the following relational expressions.

iv. The parameter is updated by the following equation using the estimated amount of step ii.
Here, λ is an update width.

v. The log likelihood lnL (k + 1) is calculated using the (k + 1) th update parameter, and the processes from (ii) to (iV) are repeated until the likelihood converges. As the convergence condition, a sufficiently small threshold is set such that lnL (k + 1) −lnL (k) <threshold. Α1 ^(k) , α2 ^(k) , and β ^(k) at the time of convergence are used in calculating the predicted value.

C) Calculation of predicted value The predicted value of household appliance power demand is calculated | required with the following processes.
Based on the estimated α1, α2, and β, the state when X ′ (t) is given in each time mesh is determined, and the predicted value is calculated by the following equation.
The average demand amount aveE _H (H, P, t) is added to this to obtain a predicted value of the home appliance power demand amount in each plan mesh.

13.2.1.5.12. Calculation of Prediction Value Swing Width Prediction error with past results is calculated for each planned mesh in the prediction period by the following process. There are as many prediction error values as there are past results in each planning mesh. For all plan meshes in the forecast period: A) The day to which the plan mesh belongs is the forecast date.
B) All dates in the same date category as the forecast date are acquired from the actual period.
C) For the plan mesh, the difference between the home appliance power demand (actual) and the home appliance power demand (predicted value) in the same time zone on the day acquired as the actual is calculated, and the prediction error in the plan mesh To do.

13.2.1.5.13. Demand scenario creation and probability of occurrence A demand scenario is created by the following process.
A) The percentile section occupied by each home appliance power demand scenario is acquired from the property file.
Example of section designation) 0.0, 5.0, 25.0, 75.0, 95.0, 100.0
B) The representative percentile value is calculated from the percentile section occupied by each home appliance power demand scenario. The representative percentile value is the midpoint of the boundary of the section.
Examples of representative percentile values) 2.5, 15.0, 50.0, 85.0, 97.5
C) The probability of occurrence of a home appliance power demand scenario is calculated from the percentile section. The occurrence probability of the home appliance power demand scenario is the absolute value of the difference between the tile values at both ends of each percentile section.
Example of probability of occurrence of demand for home appliance electricity scenario) 5.0, 20.0, 50.0, 20.0, 5.0
D) The time transition of the representative percentile point of the error is obtained for each planning mesh in the prediction period by the process of 13.2.1.5.14.
E) Obtain the time transition of the difference value for each representative percentile point.
F) Add the time transition of the difference value of each percentile point to the predicted value to obtain a home appliance power demand scenario corresponding to each representative percentile value.
G) The probability of occurrence corresponding to the representative percentile value is assigned to each home appliance power demand scenario.

13.2.1.5.14. Obtaining the transition of error percentile points The transition of error percentile points is obtained by the following process. For all planning meshes during the forecast period,
A) The prediction date to which the plan mesh belongs is acquired.
B) From the scenario table, obtain all prediction errors of each plan mesh on the past i-th forecast date for the home appliance power demand scenario (up to the past two years).
C) For each planning mesh i. A point with a prediction error of 0 is a 50% percentile point.
ii. Each percentile point is obtained from the actual distribution of prediction errors.
By the above processing, the percentile points of each error value for each planned mesh are obtained. By sorting this by the time of the planning mesh, the time transition of the percentile point of each error is obtained.

13.1.5.15. Save to DB For multiple created consumer electronics demand demand scenarios, with scenario error, percentile value, probability of occurrence and scenario classification as 1, scenario ID, target date, forecast date, forecast start date, time difference, Predicted values and prediction errors are stored in the scenario table.

13.2.2. EV electric power demand scenario creation 13.2.1. Functional Overview FIG. 128 is a diagram illustrating an IO related diagram of the EV power demand scenario creation function. From the EV power demand record data and the time-series EV travel reservation data, the EV power demand scenario in the designated house / period, the percentile value / probability thereof, and the like are stored in the database. This section refers to the EV power demand scenario creation function for EV homes. In a house without an EV, an EV power demand scenario with a demand value of 0 and an occurrence probability of 0 is always created regardless of past history and other input values.

13.2.2.2.2 Arguments and Return Value 13.2.2.2.1 Arguments FIG. 129 is a diagram illustrating arguments of the EV power demand scenario creation function.

13.2.2.2.2 Return Value FIG. 130 is a diagram illustrating a return value of the EV power demand scenario creation function.

13.2.2.2.3 Error Check Specification FIG. 131 is a diagram illustrating an error check specification of the EV power demand scenario creation function. When it corresponds to the error check, the process is interrupted and a return value is returned to the main process as a process failure. Refer to 13.2.2.2.2 Return value for details.

13.2.2.4. DB Editing Specification Acquisition and storage of values from the database in the EV power demand scenario creation function will be described.
13.2.2.4.1. Acquired Data List FIG. 132 is a diagram illustrating acquisition of values from the database.

13.2.2.4.2. List of Stored Data FIG. 133 is a diagram illustrating storage of values in the database.

13.2.2.4.3. Update data list This function does not update the database.

13.2.2.5. Processing details 13.2.2.5.1. Processing Flow FIG. 134 is a diagram for explaining an EV power demand scenario creation processing flow. In the EV power demand scenario creation process, the EV power demand scenario and its occurrence probability are obtained by the flow of FIG. 134 and stored in the database. In this process, the prediction error is not calculated.

13.2.2.5.2. Definition of season Season consists of three seasons: winter, summer, spring and autumn. Each period of winter, summer, spring and autumn is obtained from parameter information.

13.2.2.5.3. Determination of Forecast Period / Actual Period The forecast period means a period in which the EV power demand value for each plan mesh is subject to forecast. The prediction period is from the control rule creation period start date and time given as an argument to the control rule creation period end date and time (including the start point and not the end point). The actual period is a period having data used for learning the prediction model. The performance period is a set of days having EV power demand results in all plan meshes in the EV power demand performance table (the days do not have to be consecutive). However, data from past 2 years will not be used.

13.2.2.5.4. EV Electric Power Demand Scenario Creation from EV Usage Reservation An EV electric power demand scenario is created from an EV travel reservation designated by the user by the following processing.
A) From the EV electric power demand record, the electric power demand when the past EV usage is not 0 is acquired, and the average value is obtained. However, data from more than 2 years is excluded.
B) The use start time of EV travel reservation in the prediction period is obtained.
C) For all the predicted days, the past travel time on the same day of the week as the predicted date is acquired from the EV travel history, and the 90% tile point is obtained. However, data from more than 2 years is excluded.
D) A value obtained by adding the travel time obtained in C to the use start time obtained in B is obtained. This is the use end time. The period from the use start time to the use end time is called an EV travel reservation time zone.
E) For each plan mesh during the prediction period, the average value of the EV power demand obtained in A is given when included in the EV travel reservation time zone, and 0 is given otherwise.
F) The values in 10-minute units created in E) are recalculated in units of planned mesh.
The EV power demand scenario is made with the value of the planned mesh unit created by the above processing. If there is no EV travel reservation in the forecast period, no EV power demand scenario is created from the EV travel reservation.

13.2.2.5.5. EV Electric Power Demand Scenario Creation from EV Travel Performance The EV power demand scenario is created from the past EV travel performance by the following processing.
A) The first day of the forecast period is acquired and this is set as the start day. Further, the number of days over which the forecast period spans is defined as the forecast period days.
B) Find the season of the starting day.
C) The number N of EV power demand scenarios and the number M of EV power demand scenario abnormal value exclusions are acquired from the parameter information.
D) Among the dates with past EV power demand results, the period of the same day as the start day of the week, the period of zero or more EV power demand results (no data loss) for seven consecutive days from the same season, and The maximum (N-1 + M) periods in which no holidays are included in the consecutive forecast period days are acquired.
E) Obtain the transition of EV power demand for all the periods selected in the process of D.
F) When there is a weekday and a holiday in the forecast period, the day of the week that is a weekday and a holiday is specified.
G) Obtain all EV power demand on the day of acquisition on all days that fall on weekdays and holidays among the dates with past EV power demand results.
H) In the transition of the EV power demand obtained in E, the EV power demand corresponding to the day of the week specified in F is replaced with the EV power demand obtained in G. The replacement at this time is assumed to be random sampling by restoration extraction from F daily unit data.
I) For each day of the acquired period, obtain the total EV power demand record for one day. At this time, when the number of acquired periods is less than N-1 + M, the total EV power demand results for one day are ranked in ascending order, and periods including days that are Nth or less are excluded.
An EV power demand scenario is assumed by the transition of the power demand amount in units of the planned mesh created by the above processing.

13.2.2.5.6. Acquisition of EV electric power demand results EV electric power demand results are accumulated as aggregated values every 10 minutes. The planning mesh is variable and is obtained from parameter information. In the processing in this module, the actual value of EV power demand for each planned mesh is required. Therefore, when the total width of the actual value is different from the planned mesh width, the EV power demand actual value is obtained by recalculating the EV power demand actual value in the EV power demand actual table with the total width of the plan module. Here, it is assumed that the planned mesh width is longer than or equal to the total width of the actual values.

13.2.2.5.7. Assignment of EV power demand scenario occurrence probability An occurrence probability is assigned to an EV power demand scenario by the following process. The number of EV power demand scenarios created from the EV driving performance is N _h = N−1.
A) When there is an EV travel reservation during the prediction period The occurrence probability P _reg is given to the EV power demand scenario created from the EV travel reservation.
The occurrence probability (1-P _reg ) / N _h is _assigned to the EV power demand scenario created from the past results.
B) If you do not have the EV travel reservation during the forecast period in EV power demand scenarios created from past experience to give the probability 1 / N _h.

13.2.2.5.8. Saving to DB For multiple EV power demand scenarios created, scenario ID, target date / time, forecast value, forecast start date / time difference / prediction in scenario master table with prediction error / percentile value / occurrence probability and scenario classification as 2 Store values and prediction errors in the scenario table.

13.2.3. Creation of heat demand scenario 13.2.1. Functional Overview FIG. 135 is a diagram illustrating a heat demand scenario creation function IO related diagram. The occurrence probability of the heat demand scenario / prediction error / heat demand scenario in the specified house / period is obtained from the heat demand record data and the temperature of the weather information record / forecast, and stored in the database.

13.3.2.2. Arguments and return value 13.2.3.2.1. Arguments FIG. 136 is a diagram illustrating arguments of the heat demand scenario creation function.

13.2.3.2.2. Return Value FIG. 137 is a diagram illustrating a return value of the heat demand scenario creation function.

13.2.3. Error Check Specification FIG. 138 is a diagram illustrating an error check specification of the heat demand scenario creation function. When it corresponds to the error check, the process is interrupted and a return value is returned to the main process as a process failure. For details, refer to 13.2.3.2.2 Return value.

13.2.3.4. DB editing specification The acquisition and storage of values from the database in the heat demand scenario creation function will be described.
13.2.3.4.1. Acquired Data List FIG. 139 is a diagram illustrating acquisition of values from the database.

13.2.3.4.2. List of Stored Data FIG. 140 is a diagram illustrating an example of storing values in the database.

13.2.3.4.3. Update data list This function does not update the database.

13.2.3.5. Processing details 13.2.3.5.1. Processing Flow FIG. 141 is a diagram illustrating a heat demand scenario creation processing flow. In the heat demand scenario creation process, the occurrence probability of the heat demand scenario, prediction error, percentile value, and heat demand scenario is obtained by the following flow and stored in the database.

13.2.3.5.2. Definition of date category The date category is defined as follows.
For holidays (3 categories): Season (3 categories)
If it is not a holiday (21 categories): Day of the week (7 categories) x Season (3 categories)
Seasons consist of three categories: winter, summer, autumn and autumn. Each period of winter, summer, spring and autumn is obtained from parameter information. There are 7 categories: Sunday, Tuesday, Wednesday, Thursday and Friday. Whether or not the target day is a holiday is determined based on whether or not a record corresponding to the day exists in the calendar table.
(Example) 2012/5/3 → Holidays / Spring / Autumn
2012/12/1 → not a national holiday / Saturday / winter

13.2.3.5.3. Determination of Forecast Period / Actual Period The forecast period means a period in which the heat demand value for each planned mesh is targeted for prediction. The prediction period is from the control rule creation period start date and time given as an argument to the control rule creation period end date and time (including the start point and not the end point). The actual period is a period having data used for learning the prediction model. The result period is a set of days having the heat demand results in all the planning meshes in the heat demand result table (the days do not need to be consecutive). However, data from past 2 years will not be used.

13.2.3.5.4. Creation of forecast value The forecast value of the amount of heat demand in the forecast period is created by the following process.
A) A date category is assigned to all dates in the actual period / forecast period.
B) For all dates in the forecast period i. Get all dates that belong to the same date category as the forecast date from the actual period. This is called a category performance date.
ii. Depending on the number of category actual values, determine the actual date that will be used to create the predicted value for:
iii. From the dates acquired by the above processing, the dates to be counted are selected as follows.
A) If the number of days in the category results is 1 day or more and within X days, all the category performance days are counted. Note that the value of X is obtained from the parameter information.
A) When the number of days of category performance is X + 1 days or more, the date up to the top X in the following value C among the category performance days is counted.
The maximum and minimum temperatures in the past data are the maximum and minimum values of the temperature for each time zone in the weather information performance table. The maximum and minimum temperatures on the forecast date are obtained from the weather information forecast table.
iv. The average value of the hot water temperature in each plan mesh is calculated for the day selected in (iii), and this is used as the predicted value of heat demand.
C) The prediction value of each plan mesh in the forecast period is arranged in ascending order of the time of the plan mesh, and is used as the heat demand forecast value in the forecast period.

13.2.3.5.5. Acquisition of actual heat demand values Actual heat demand values are accumulated as aggregated values every 10 minutes. The planning mesh is variable and is obtained from parameter information. In the processing in this module, the actual value of the heat demand for each planned mesh is required. Therefore, when the total width of the actual value and the planned mesh width are different, the actual heat value is obtained by recalculating the actual value in the heat demand actual table with the total width of the planning module. Here, it is assumed that the planned mesh width is longer than or equal to the total width of the actual values.

13.2.3.5.6. Calculation of Prediction Value Swing Width Prediction error with past results is calculated for each planned mesh in the prediction period by the following process. There are as many prediction errors as the number of past results in each planning mesh. For all days in the forecast period (forecast date) A) Get all dates in the same date category as the forecast date from the actual period.
B) Calculate the difference between the heat demand (actual) and the heat demand (predicted value) in the same time zone on the date acquired as the actual for all the planned meshes within the predicted date (difference value for the past date) Is obtained).

13.2.3.5.7. Creation of heat demand scenario and assignment of occurrence probability Heat demand scenario is created by the following process.
A) The percentile section occupied by each heat demand scenario is acquired from the property file.
Example of section designation) 0.0, 5.0, 25.0, 75.0, 95.0, 100.0
B) The percentile point is calculated from the percentile section occupied by each heat demand scenario. The percentile point is the midpoint of the boundary of the section.
Example of calculating percentile points) 2.5, 15.0, 50.0, 85.0, 97.5
C) The occurrence probability of the heat demand scenario is calculated from the section occupied by each heat demand scenario. The occurrence probability of the heat demand scenario is the absolute value of the difference between the tile values at both ends of each percentile section.
Example of heat demand scenario occurrence probability) 5.0, 20.0, 50.0, 20.0, 5.0
D) The time transition of the error value corresponding to the percentile point is obtained for each planning mesh in the prediction period by the process of 13.2.3.5.8.
E) Obtain the time transition of the difference value for each percentile point.
F) Add the time transition of the difference value of each percentile point to the predicted value to obtain the heat demand scenario of each percentile point.
G) A probability of occurrence corresponding to the percentile point is given to the heat demand scenario of each percentile point.

13.2.3.5.8. Calculation of transition of error percentile point A heat demand scenario is created by the following process.
A) Percentile section that each heat demand scenario occupies is acquired from the property file. Example of specifying section) 0.0, 5.0, 25.0, 75.0, 95.0, 100.0
B) The representative percentile value is calculated from the percentile section occupied by each heat demand scenario. The representative percentile value is the midpoint of the boundary of the section.
Examples of representative percentile values) 2.5, 15.0, 50.0, 85.0, 97.5
C) The occurrence probability of the heat demand scenario is calculated from the percentile section. The occurrence probability of the heat demand scenario is the absolute value of the difference between the tile values at both ends of each percentile section.
Example of heat demand scenario occurrence probability) 5.0, 20.0, 50.0, 20.0, 5.0
D) The time transition of the representative percentile point of the error is obtained for each planning mesh in the prediction period by the process of 13.2.1.5.14.
E) Obtain the time transition of the difference value for each representative percentile point.
F) The time transition of the difference value of each percentile point is added to the predicted value to obtain a heat demand scenario corresponding to each representative percentile value.
G) A probability of occurrence corresponding to the representative percentile value is assigned to each heat demand scenario.

13.2.3.5.9. Saving to DB For multiple created heat demand scenarios, scenario ID, target date / time, forecast value, forecast start date / time difference, forecast value in the scenario master table with forecast error, percentile value, occurrence probability and scenario classification as 3. Store the prediction error in the scenario table.

13.3. Supply prediction function 13.3.1. PV supply scenario creation function 13.3.1.1. Functional Overview FIG. 142 is a diagram illustrating a PV supply scenario creation function IO related diagram. Obtain the PV supply scenario / prediction error / probability of PV supply scenario in the specified house / period from the PV supply record data and the temperature / weather / insolation amount of the weather information record, and the temperature / weather of the weather information forecast. Save them in the database.

13.3.1.2. Arguments and return value 13.3.1.2.1. Arguments FIG. 143 is a diagram illustrating arguments of the PV supply scenario creation function.

1.3.1.2.1.2.2. Return Value FIG. 144 is a diagram illustrating a return value of the PV supply scenario creation function.

13.3.1.2.3. Error Check Specification FIG. 145 is a diagram illustrating an error check specification of the PV supply scenario creation function. When it corresponds to the error check, the process is interrupted and a return value is returned to the main process as a process failure. For details, refer to 13.3.1.2.2 Return value.

13.3.1.3. DB Editing Specification Acquisition and storage of values from the database in the PV supply scenario creation function will be described.
13.3.3.1. Acquired Data List FIG. 146 is a diagram illustrating acquisition of values from the database.

13.3.1.3.2. List of Stored Data FIG. 147 is a diagram illustrating storage of values in the database.

13.3.1.3. Update data list This function does not update the database.

13.3.1.4. Processing details 13.3.4.1. Process Flow FIG. 148 is a diagram for explaining a PV supply scenario generation process flow. In the PV supply scenario creation process, PV supply scenario, prediction error, percentile value, and probability of occurrence of PV supply scenario are obtained by the following flow and stored in a database.

13.3.1.4.2. Determination of Forecast Period / Actual Period The forecast period means a period in which the PV supply value for each planned mesh is subject to prediction. The prediction period is from the control rule creation period start date and time given as an argument to the control rule creation period end date and time (including the start point and not the end point). The actual period is a period having data used for learning the prediction model. The performance period is a set of days having a performance in all the planning meshes in the PV supply performance table (the days do not have to be consecutive). However, data from past 2 years will not be used.

13.3.1.4.3. Obtaining predicted temperature change Obtain the predicted temperature change by the following process.
A) When there is hourly weather forecast information The latest predicted temperature for every hour (= for each planned mesh) is acquired from the hourly weather information forecast table.
B) When there is only weather forecast information by date For the specified forecast date,
i. Get the latest forecast maximum temperature and forecast minimum temperature on the forecast date.
From the daily weather information forecast table, the highest temperature and the lowest temperature on the forecast day are acquired and set as the predicted highest temperature and the lowest forecast temperature.
ii. From the performance period, obtain the highest and lowest temperature results for all days and calculate the following values.
iii. Select as many N days as possible in ascending order of the above values, obtain the temperature results for that day from the weather information results table, and calculate the average value for each planned mesh. This value is used as the predicted temperature transition for the day.

13.3.1.4.4. Parameter Estimation The temperature coefficient γ _s θ (H) of power generation efficiency and the temperature constant γ _s0 (H) of power generation efficiency are estimated by the following processing.
A) The PV supply amount record s (H, t) is acquired from the PV supply record table.
B) Acquire the temperature record θ (t) and the solar radiation record 1 (H, t) from the elephant information record table.
C) Using s _i (H, t), θ _i (t), and l _i (H, t) as individual PV supply amount results, temperature results, and solar radiation results data, the following values are the smallest γ ^ _S θ, γ ^ _s0 are obtained.

13.3.1.4.5. Creation of predicted value The predicted value s ′ (H, t) of the PV supply amount is obtained by the following process.
A) Acquire the predicted temperature and solar radiation results for each planned mesh during the forecast period.
B) Using the following formula, obtain the PV supply amount of each planned mesh using γ ^ _s θ, γ ^ _s0 obtained by the parameter estimation process and the predicted temperature / insolation results.
This value is used as a predicted value of the PV supply amount.

13.3.1.4.6. Calculation of fluctuation of prediction value Prediction error is calculated by the following process. The difference between the PV supply predicted value and the PV supply actual value is taken for each plan mesh for all days (predicted date) in the prediction period. This is used as a prediction error.

13.3.1.4.7. Creation of PV supply scenario and assignment of occurrence probability PV supply scenario is created by the following process.
A) The percentile section occupied by each PV supply scenario is acquired from the property file.
Example of section designation) 0.0, 5.0, 25.0, 75.0, 95.0, 100.0
B) The percentile point is calculated from the percentile section occupied by each PV supply scenario. The percentile point is the midpoint of the boundary of the section.
Example of calculating percentile points) 2.5, 15.0, 50.0, 85.0, 97.5
C) The probability of occurrence of the PV supply scenario is calculated from the section occupied by each PV supply scenario. The occurrence probability of the PV supply scenario is the absolute value of the difference between the tile values at both ends of each percentile section.
Example of PV supply scenario occurrence probability) 5.0, 20.0, 50.0, 20.0, 5.0
D) The time transition of the error value corresponding to the percentile point is obtained for each planning mesh in the prediction period by the process of 13.3.1.4.8.
E) Obtain the time transition of the difference value for each percentile point.
F) The time transition of the difference value of each percentile point is added to the predicted value to obtain a PV supply scenario of each percentile point.
G) A probability of occurrence corresponding to the percentile point is given to the PV supply scenario of each percentile point.

13.3.1.4.8. Calculation of transition of error percentile point A PV supply scenario is created by the following process.
A) The percentile section occupied by each PV supply scenario is acquired from the property file.
Example of section designation) 0.0, 5.0, 25.0, 75.0, 95.0, 100.0
B) A representative percentile value is calculated from the percentile section occupied by each PV supply scenario. The representative percentile value is the midpoint of the boundary of the section.
Examples of representative percentile values) 2.5, 15.0, 50.0, 85.0, 97.5
C) The occurrence probability of the PV supply scenario is calculated from the percentile section. The occurrence probability of the PV supply scenario is the absolute value of the difference between the tile values at both ends of each percentile section.
Example of PV supply scenario occurrence probability) 5.0, 20.0, 50.0, 20.0, 5.0
D) The time transition of the representative percentile point of the error is obtained for each planning mesh in the prediction period by the process of 13.2.1.5.14.
E) Obtain the time transition of the difference value for each representative percentile point.
F) A PV supply scenario corresponding to each representative percentile value is obtained by adding the time transition of the difference value of each percentile point to the predicted value.
G) An occurrence probability corresponding to the representative percentile value is assigned to each PV supply scenario.

13.3.1.4.9. Saving to DB For multiple PV supply scenarios that have been created, the scenario ID, target date / time, prediction value, prediction start date / time difference / prediction value in the scenario master table with prediction error / percentile value / occurrence probability and scenario classification as 3 Store the prediction error in the scenario table.

13.4. Optimal supply and demand planning function 13.4.1. Integrated scenario creation 13.4.1.1. Functional Overview FIG. 149 is a diagram illustrating an IO related diagram of the integrated scenario creation function. The home appliance power demand scenario, EV power demand scenario, heat demand scenario, and PV supply scenario data are integrated to create integrated scenario data.

13.4.1.2. Argument and return value 13.4.1.2.1. Arguments FIG. 150 is a diagram illustrating arguments of the integrated scenario creation function.

13.4.1.2.2.2. Return Value FIG. 151 is a diagram illustrating a return value of the integrated scenario creation function.

13.4.1.3. Error Check Specification FIG. 152 shows the error check specification of the integrated scenario creation function. When it corresponds to the error check, the process is interrupted and a return value is returned to the main process as a process failure. For details, refer to 13.4.1.2.2.2 Return value.

13.4.1.4. DB Editing Specification Acquisition and storage of values from the database in the integrated scenario creation function will be described.
13.4.1.4.1. Acquired Data List FIG. 153 is a diagram illustrating acquisition of values from the database.

13.4.1.4.2. List of Stored Data FIG. 154 is a diagram illustrating an example of storing values in the database.

13.4.1.4.3. Update data list This function does not update the database.

13.4.1.5. Processing details 13.4.1.5.1. Processing Flow FIG. 155 is a diagram for explaining the integrated scenario creation processing flow. In the integrated scenario creation process, the probability of occurrence of an integrated scenario / integrated scenario is obtained by the following flow and stored in a database. The integrated scenario is a total of four time series values obtained by selecting each scenario from the consumer electronics power demand scenario group, EV power demand scenario group, heat demand scenario group, and PV supply scenario group, and the accompanying information. Point to.

13.4.1.5.2. Creating a scenario combination A demand / supply scenario combination is created by the following process.
A) The scenario ID of the home appliance power demand scenario, EV power demand scenario, heat demand scenario, and PV supply scenario corresponding to the control rule creation process ID is acquired from the scenario table.
B) Select one scenario each for the household electric power demand scenario, EV power demand scenario, heat demand scenario, PV supply scenario, and create a combination of four scenario IDs. This combination of scenario IDs is created for all patterns.

13.4.1.5.3. Calculation of probability of occurrence of integrated scenario The probability of occurrence of all integrated scenarios is calculated by the following processing.
A) Predictive value, prediction error, probability of occurrence, percentile value, home appliance power demand scenario table, EV power corresponding to each scenario ID of the integrated scenario for the home appliance power demand scenario, EV power demand scenario, heat demand scenario, PV supply scenario Obtained from the demand scenario table, heat demand scenario table, and PV supply scenario table.
B) Create a correlation matrix of prediction errors of household electric power demand scenario, heat demand scenario, and PV supply scenario.
C) FIG. 156 shows an example of processing. The correlation matrix is cleaned by processing this R sample.
D) For all integration scenarios,
i. Through the following processing, the probability of simultaneous occurrence of a combination of scenarios constituting each scenario group is calculated for three scenarios, namely, a home appliance power demand scenario, a heat demand scenario, and a PV supply scenario.
A) Assuming a standard normal distribution with an average of 0 and a variance of 1, a rank corresponding to the tile value of each scenario is calculated.
B) Obtain the probability of simultaneous occurrence of three scenarios using a three-dimensional normal density function. FIG. 157 shows an implementation example using R.
ii. The occurrence probability of the EV scenario is multiplied by the occurrence probability of the three scenarios obtained previously to obtain the occurrence probability P _S of the integrated scenario S.

13.4.1.5.4. Selection of integration scenario and standardization of occurrence When the created integration scenario exceeds the maximum number N of integrated scenarios, the top N are selected in descending order of occurrence probability and used as the output of this module. For the selected N integration scenarios, the occurrence probability P ^* _S of the integration scenario S standardized by the following equation is obtained, and this is set as the final occurrence probability.
The maximum integrated scenario number N is acquired from the parameter information.

13.4.1.5.5. Saving to DB Stores the electricity demand forecast value / EV power demand forecast value / heat demand forecast value / PV supply forecast value and the occurrence probability P ^* _S for it in the unified scenario table for up to N integrated scenarios created. To do.

13.4.2. Optimization plan creation 13.4.2.1. Functional Overview FIG. 158 is a diagram illustrating an IO related diagram for creating an optimization plan. Create various optimization plans from the integrated scenario.

13.4.2.2. Arguments and return value 13.4.2.2.1. Arguments FIG. 159 is a diagram illustrating arguments of the optimization plan creation function.

13.4.2.2.2. Return Value FIG. 160 is a diagram illustrating a return value of the optimization plan creation function.

13.4.2.3. Error Check Specification FIG. 161 is a diagram illustrating an error check specification of the optimization plan creation function. When it corresponds to the error check, the process is interrupted and a return value is returned to the main process as a process failure. Refer to 13.4.2.2.2 Return value for details.

13.4.2.4 DB Editing Specification Acquisition and storage of values from the database in the optimization plan creation function will be described.
13.4.2.4.1. Acquired Data List FIGS. 162 to 166 are diagrams illustrating acquisition of values from the database.

13.4.2.4.2. List of Stored Data FIGS. 167 to 169 are diagrams illustrating an example of storing values in the database.

13.4.2.2.3. Update data list This function does not update the database.

13.4.2.5. Processing details 13.4.2.1. Processing Flow FIG. 170 is a diagram illustrating a processing list. In the optimization plan creation process, the following A) to F) processes are executed in order to perform the optimization calculation and calculate the optimal operation plan for the EV storage battery, stationary battery and FC. At the end of the process, a process of saving the calculation result in the database is performed.

A) Initialization processing Constants that need to be set for initialization are obtained from the database as described in 13.4.2.1. The total number of acquired scenario IDs is N, and each ID is numbered 1, 2,. Further, assuming that the total number of meshes in one span to be processed is T, each mesh is numbered 1, 2,... In order to indicate the presence or absence of EV connection, a value that is 1 when EV is connected (not used) in period t of integration scenario i, and 0 otherwise is set as δ _EVi (t). When the optimization mode is emergency mode, the system is disconnected from the grid, so z _max ⁻ (t) and z _max ⁺ (t) are fixed to 0, and EV is also connected to use PV power. E _Vi (t) is set to 0, and δ _EVi (t) is set to 1.

B) Variable Creation FIGS. 171 to 173 are diagrams illustrating a list of variables necessary for defining an optimization model. When defining variables, specify the "type" and range as specified in the table. There are two types of types, “continuous” and “0-1”, and an upper limit and a lower limit are specified for the range. When there is no need to specify an upper limit or a lower limit, “−” is indicated in the table. Each T + 1 variable is generated. For example, if the variable of backward flow power coulometric, to generate a T + 1 or to z ⁺ (T + 1) from the z ⁺ (1). If z ⁺ (t), it becomes a variable representing the reverse tide flow rate in the t period (from the beginning to the t th mesh period) of the planned span. In general, when T variables are expressed as x (1), x (2),..., X (T), variable definition in Gurobi is performed by Model.addVar (). FIG. 174 shows an example described in Python.

 Here, vtype specifies the type. Specify GRB.CONTINUOUS if continuous, or GRB.BINARY if 0-1. If the upper and lower limits of the variable value range are specified, specify the lower limit with lu and the upper limit with ub. This is unnecessary if the upper and lower limits of the range are not specified, but it is also possible to write lb = -GRB.INFINITY or ub = GRB.INFINITY to explicitly indicate that the upper and lower limits are not specified. .

C) Constraint Condition Creation FIGS. 175 to 177 are diagrams illustrating constraint conditions. In general, when N variables are expressed as x (1), x (2),..., X (T), a linear equality or inequality is represented by N constants a as coefficients of each variable. (1), a (2), ..., a (T) and the constant b on the right side
It is determined.

The constraint conditions in Gurobi are defined by defining the linear expressions on the right and left sides with LinExpr () and adding them to the model with model.addConstr (). FIG. 178 shows an example implementation in Python when a linear equation is generated and added to the model. In this example, a pair of a coefficient a (t) and a variable x (t) is added to the expression on the left side of addTerms (a [t], x [t]), and a constant b is added using addConstant (b). To add constraints to the model, use model.addConst () and specify the primary expression on the left side for lhs and the right side for rhs. The relationship between the right side and the left side is specified by sense. In the case of an equal sign, sense = "=", and in the case of an inequality sign, sense = "<=" or sense = ">=" is specified depending on the direction. When the optimization mode is emergency, the charge amount becomes variable from a fixed value, so the 13th constraint condition in the above table is
It is necessary to change.

D) Objective Function Creation FIG. 179 is a diagram illustrating a linear expression constituting the objective function. As shown in the table below, the objective function is composed of the following linear expressions corresponding to costs, number 1 to number 5, and the risk amount (number 6) calculated for each integration scenario i. The objective function is set as follows using the occurrence probability p _i of the integrated scenario i and the risk parameter θ.
Note that setting the objective function in Gurobi is done by setting a linear expression of the objective function in expr and adding it to the model with model.setObjective (). FIG. 180 shows an implementation example using Python. In this example, GRB.MINIMIZE is specified to perform minimization.

E) Implementation of Optimization Calculation FIGS. 181 and 182 are diagrams illustrating risk parameters and variables corresponding to output items. Generate an objective function that sets each of the three risk parameters θ _S , θ _L, and θ _U (see 14.2 Initial Parameters for setting values) obtained from the optimization engine parameter table as the risk parameter θ of the objective function. Then, optimization processing is executed for each objective function (optimization processing is executed three times). FIG. 181 and FIG. 182 exemplify setting values of risk parameters used to generate an optimization variable and an optimization objective function used to set the value for each output item.

  In order to execute optimization calculation with Gurobi, model.optimize () is used after setting constraint equations and objective functions as models. Thereby, an optimal solution is calculated by an algorithm capable of obtaining a global optimal solution. After optimization, optimized values are set for each variable. FIG. 183 shows an example of obtaining the variable x [t] using Python.

F) Result storage Each item shown in FIGS. 181 and 182 is stored in the table shown in the list of stored data in 1.4.2.4.2.

13.4.3. Control rule generation 13.4.3.1. Functional Overview FIG. 184 is a diagram illustrating an IO diagram of the control rule creation function. A control rule is created from each optimized plan data.

13.4.3.2 Arguments and return value 13.4.3.2.1. Arguments FIG. 185 is a diagram illustrating arguments of the optimization plan creation function.

13.4.3.2.2. Return Value FIG. 186 is a diagram illustrating a return value of the optimization plan creation function.

13.43.3. Error Check Specification FIG. 187 is a diagram illustrating an error check specification of the optimization plan creation function. When it corresponds to the error check, the process is interrupted and a return value is returned to the main process as a process failure. For details, see 13.4.3.2.2 Return value.

13.4.3.4. DB Editing Specification Acquisition and storage of values from the database in the optimization plan creation function will be described.
13.4.3.4.1. Acquired Data List FIG. 188 is a diagram illustrating acquisition of values from the database.

13.4.3.4.2. List of Stored Data FIG. 189 is a diagram illustrating storage of values in the database.

13.4.3.4.3. Update data list This function does not update the database.

13.4.3.5. Processing details 13.4.3.5.1. Processing Flow FIG. 190 is a diagram for explaining a control rule generation processing flow. In the control rule generation process, the control rules for each device are obtained from the optimization plan of the EV storage battery / stationary storage battery / fuel cell according to the following flow and stored in the database.

13.4.3.5.2. Device Control State FIG. 191 is a diagram illustrating control on a device. Specify the following control status for each device.

13.4.3.5.3. Basic State of Device FIG. 192 is a diagram illustrating the basic state of the device (when the system power supply is connected). FIG. 193 is a diagram illustrating a basic state of the device (when disconnected from the system power supply). The device state in each plan mesh derived by the optimization plan is called a basic state. In the EV storage battery and the stationary storage battery, discharging is performed when the amount of discharging power is positive in the planned mesh, charging is performed when the amount of charging power is positive, and standby is performed when both values are 0 (meaning a state where neither charging nor discharging is performed). ). The fuel cell is activated when the fuel cell power generation amount is positive, and stopped when it is zero.

13.4.3.5.4. Calculation of the management width of the storage battery in the planned mesh The calculation width of the storage battery in each planned mesh is calculated. Each house is provided with either EV (storage battery) or stationary storage battery. Here, the management width of the devices installed in each house will be mentioned. It is assumed that the upper limit of the management width _changes at a certain rate in the planned mesh, and the upper limit at the elapse of t minutes is a straight line connecting the upper limit value V _1max of the planned mesh and the upper limit value V _2max of the next planned mesh. And Similarly, the lower limit of the control range, the straight line connecting the lower limit V _2min lower limit V _1min and next plan mesh of the plan mesh and the lower limit value in the control mesh. When the planned mesh width is 60 minutes, the upper and lower limits v _max (t) and v _min (t) of the management width when t minutes elapse in each planned mesh can be obtained by the following equations, respectively.

13.4.3.5.5. Rule generation processing (when grid power is connected)
In the case of an EV installation house, the process A is performed, and in the case of a stationary storage battery setting house, the process B is performed.
A) Creation of EV storage battery control rules (for EV installation houses)
An EV storage battery control rule is created by the following process.
i. Acquisition of EV storage battery charge / discharge plan The EV storage plan upper limit value / EV storage plan lower limit value is acquired from the EV storage plan table, and the EV discharge capacity plan / EV charge capacity plan is acquired from the EV power plan table.
ii. Creation of basic state A basic state for each planned mesh is created from an EV discharge power plan and EV charge power plan for each planned mesh.
iii. Acquire the control mesh width The control mesh width is acquired from the parameter information, and the planned mesh is divided into control mesh units. However, when the control mesh width exceeds the planned mesh width, it is set equal to the planned mesh width.
iv. Management width calculation in control mesh The management width in each plan mesh is calculated by the processing of 13.4.3.5.4.

v. Control Rule Creation Method Within each control mesh, a control rule is created according to the following rules.
(A) When the basic state is discharging (a) When the measured value of the amount of stored EV is within the management width or exceeds the upper limit of the management width, the EV storage battery is discharged.
(B) When the measured value of the EV storage amount falls below the lower limit of the management width, the EV storage battery enters a standby state.
(A) When the basic state is charging (a) When the measured value of the EV storage amount is within the management range or exceeds the lower limit of the management range, the EV storage battery is charged.
(B) When the measured value of the EV storage amount exceeds the upper limit of the management width, the EV storage battery enters a standby state.
(C) When the basic state is standby (a) The EV storage battery is in a standby state regardless of the management width.

  FIG. 194 shows an example of the EV storage battery storage plan value and the basic state.

vi. Control Rule Format FIG. 195 is a diagram illustrating a storage battery control rule created from FIG. 194. The control rule is described in an if-then format. In the if section, a time or a magnitude comparison expression is described, and the operation target device and the operation state of the device are described after then. The general formula is as follows.
・ Time activation: if YYYYMMDDhhmm then action (object, state)
・ Comparison activation: if variable <threshold then action (object, state)

B) Creating control rules for stationary storage batteries (for homes with stationary storage batteries)
A control rule for the stationary storage battery in the control mesh is created by the following processing.
i. Acquisition of Charging / Discharging Plan for Stationary Storage Battery The LIB storage plan upper limit value / LIB storage plan lower limit value is acquired from the LIB storage plan table, and the LIB discharge capacity plan / LIB charging capacity plan is acquired from the LIB power plan table.
ii. Creation of a basic state A basic state for each planned mesh is created from a LIB discharge capacity plan / LIB charge capacity plan for each planned mesh.
iii. Acquire the control mesh width The control mesh width is acquired from the parameter information, and the planned mesh is divided into control mesh units. However, when the control mesh width exceeds the planned mesh width, it is set equal to the planned mesh width.
iv. Management width calculation in control mesh The management width in each plan mesh is calculated by the processing of 13.4.3.5.4.
v. Method for creating control rules Same as the method for creating control rules for EV storage batteries.

C) Preparation of fuel cell control rules i. Obtain FC power generation plan Obtain FC power generation plan value from FC calorific value plan table.
ii. Creation of basic state A basic state for each planned mesh is created from the FC power generation plan value for each planned mesh.
iii. Obtaining the fuel cell control mesh width The control mesh width is obtained from the parameter information, and the planned mesh is divided into control mesh units. However, when the control mesh width exceeds the planned mesh width, it is set equal to the planned mesh width.
iv. Control Rule Creation Method The basic state of each planned mesh is determined according to an optimization algorithm that has been optimized in consideration of energy loss at startup. The control rules are also as shown in the table below.

  FIG. 196 is a diagram illustrating optimization mode and fuel cell control.

13.4.3.5.6. Rule generation processing (system power off: emergency mode)
A) Creation of EV storage battery control rules (for EV storage battery installation houses)
In the emergency mode, the storage battery is set to the ON / load following mode in all the control meshes in the planned mesh, and a control rule for the EV storage battery is created.
B) Generation of control rules for stationary storage batteries (in the case of stationary storage battery installation houses)
A method for creating a stationary storage battery control rule in the emergency mode conforms to that of the EV storage battery.
C) Fuel Cell Control Rule Creation FIG. 196 is a diagram illustrating control of the fuel cell when the storage battery is in a discharged state. FIG. 197 is a diagram illustrating control of the fuel cell when the storage battery is in a charged state. As shown in these figures, the control rules of the fuel cell are created according to the following rules in all control meshes in the planned mesh.
i. When the measured value of the charged amount falls below the lower limit of the management width by ΔkWh, the output of the fuel cell is increased by the value ΔW calculated as follows. However, ΔT is the time interval of the planned mesh width, and the increased output is not set to be greater than the maximum output of FC.
ii. When the measured value of the charged amount exceeds the upper limit of the management width by ΔkWh, the output of the fuel cell is decreased by a value ΔW calculated as follows. However, ΔT is a time interval of the planned mesh width, and the output after reduction is not set to be less than the minimum output of FC.
iii. When the measured value of the stored amount is within the management range, the output of the fuel cell is not changed (the previous output is maintained).

13.4.3.5.7. Save to DB Save the created control rule in the control rule table.

14 Initial processing 14.1. Initial processing There is no initial processing in this system. Also, no special processing is required when the system is restarted.

14.2. Initial Parameters FIGS. 199 to 201 show items to be copied to the time series optimization parameter table among the setting items of the house master. When using this system, it is necessary to set the user or administrator to each item of the house master. For each item set in the house master, after accepting the control rule processing request, the settings in the house master are copied to the time series optimization parameter table, optimization parameter table, optimization engine parameter table, and a series of The control rule generation process is called.

It should be noted that the parameter values need to be set as follows depending on the presence or absence of equipment set in the house.
(A) In the case of a house without EV, the following parameters are set to 0.
・ Minimum discharge capacity of EV storage battery ・ Maximum discharge capacity of EV storage battery ・ Minimum charge capacity of EV storage battery ・ Maximum charge capacity of EV storage battery (A) For a house without a stationary storage battery, set the following parameters to 0.
・ Minimum discharge capacity of stationary storage battery ・ Maximum discharge capacity of stationary storage battery ・ Minimum charging capacity of stationary storage battery ・ Set the following parameters to 0 for homes without fixed storage battery maximum charging capacity (c) FC.
・ Minimum power generation of fuel cell ・ Maximum power generation of fuel cell

  FIG. 202 shows items to be copied to the optimization parameter table among the setting items of the house master.

  FIG. 203 shows items to be copied to the optimization engine parameter table among the setting items of the housing master. Note that the set values in the table may be changed in future simulations in order to satisfy the processing time requirement and to improve prediction accuracy.

15. Fault handling and error handling The following are examples of errors that can occur in this system.
A) Database connection / query issuance error B) R, Gurobi, etc. library error C) Other program bug error D) Optimization engine IF connection error E) HW failure error F) When the difference between the predicted value and the actual value is more than a certain value

  When the errors A to E occur, the control rule cannot be updated. As for the influence range, even if any of the above errors occurs, data inconsistency does not occur because data is managed separately for each process. When the errors A to D occur, an error occurrence message is output to the log. When the errors A to C occur, the optimization engine IF is notified that the series of processing up to the generation of the control rule has ended abnormally in addition to the previous processing. The protocol interface for notification is described in Section 11. As a result, when errors A to C occur, the caller can detect the occurrence of the error and take a workaround. When the error D occurs, the optimization engine IF cannot start the control rule generation process or cannot return a response to the optimization engine IF. By detecting connection errors on the call side even when errors D and E occur, or if there is no response after a certain period of time, without waiting for a response, we go to get the control rule of the previous generation, It is possible to take a workaround. In any case, since the control rule for one week ahead is held, control can be performed without a fatal problem if the failure is recovered within one week. Further, even when the failure is longer than one week, the same control as in the previous week can be continued by applying the control rule of the same day of the previous week. Further, when a concern that the control rule adversely affects for some reason, it is possible to turn off the control rule based on the overall optimum plan and perform fixed control with the control rule of STEP1. Regarding F, since the deviation width and the deviation rate between the predicted value and the actual value are described in the log, it is possible to immediately notice when the deviation width is a certain value or more by monitoring the log.

16. Maintenance operation Data to be backed up in this system is data in the database. A monitoring item for failure monitoring is a log file. What is output to the log is (1) that each process ended normally and its date and time, or (2) errors when connecting to the database and issuing queries, errors in the library used by R, Gurobi, etc., optimization engine An error in connection to the IF, an error due to a bug in the program, etc. has occurred and the process ended abnormally, and the date and time and the stack trace, or (3) the difference between the predicted value and the actual value. Therefore, by monitoring the log, (1) when and which process caused the error, (2) when and which process caused the delay, and (3) the prediction was significantly different from the actual results. Can be detected. However, (1) The cause of the error and the repair method, etc., (2) The cause of the processing time taking longer than the specified value and the repair method, etc. (3) The prediction is significantly different from the actual results. How to determine what is happening, the cause of the divergence and how to repair it are determined at the detailed design stage. The log file output destination and log file name are determined at the detailed design stage. The specific backup method, specific monitoring method, and operation flow are left to the operations staff.

17. System non-functional requirements 17.1. Performance Requirement Processing time is measured from the time when data acquisition from the DB is completed in each processing, and the processing time requirements obtained by accumulating the values for all the processing are defined as follows. The control rule generation process must be completed within 30 minutes for a control rule generation request for one house. This is because it is known that it takes about the same processing time from the necessity of processing the control rule generation processing for at least 16 cases per day and from past similar results. In addition, control rule generation processing must be completed within 5 hours for the control rule generation requests for 16 units issued simultaneously.

  The time from when the optimization engine I / F makes a process start request until it receives a process completion notification request depends on the network status, the performance of various DB servers, and the resource status. Not set. Therefore, it is necessary to perform measurement with an actual machine separately.

17.2. Security Security is not considered because connections from external systems are specified and limited.

17.3. Reliability The parts related to system reliability such as redundant configuration and load balance control are left to the operations personnel.

17.4. Maintainability Operations such as operation monitoring and data backup are left to the operations staff.

18. Preconditions / Constraints 18.1. Operation Restriction FIG. 204 is a diagram illustrating a precondition for system operation. This scope of application may be extended in various ways.

18.2. Preconditions for optimization There are the following preconditions for execution of optimization.
・ Each house has either EV storage battery or stationary storage battery.
Each device can be controlled independently regardless of the state of charging, discharging, stopping, etc. of other devices (for example, EV discharging / charging control is possible without depending on the operating status of the fuel cell or PV).
・ Each device operates without delay when it receives a control command, and transitions to the specified state. However, in order for the FC to transition from the stopped state to the activated state, an activation time of 1 mesh time is required.
・ Each device is connected to a house, and the generated power can be used for all consumer electronics demand.
-The power generated by each device can be reversely flowed into the system.
-The fuel cell can be operated continuously with a radiator using a radiator even when it is full. There is no upper limit to the duration of continuous operation.
・ The amount of power generated by the fuel cell is proportional to the consumed gas fuel.
・ The calorific value of the fuel cell is proportional to the consumed gas fuel.
・ The amount of heat generated by the gas boiler is proportional to the consumed gas fuel.

19. Reference material 19.1. List of Specifiable Parameters FIG. 205 to FIG. 207 show parameters that can be specified in this system as follows. These are stored in the property file as property information.

19.3. Database table list (logical design)
208 to 210 are diagrams illustrating table lists. For a table with a circle in the “intermediate data” column, a backup having the same items is created.

19.4. Estimating Number of Records by Table FIGS. 211 and 212 are diagrams illustrating examples of calculation formulas and approximate values for the number of records stored in each table. FIG. 213 is a diagram illustrating parameter values used for calculating the approximate value. The value of trial calculation 1 assumes a demonstration experiment (16 houses) conducted with the system, and the value of trial calculation 2 assumes commercialization (10,000 houses) and measures processing by shortening the span length. The value obtained by increasing the speed and the trial calculation 3 are values when the tables are partitioned in units of 7 days with respect to the trial calculation 2.

19.5. Database item list (logical design)
214 to 250 are diagrams illustrating examples of data items in each table. The annotations in FIGS. 214 to 250 are as follows.
* 1 Corresponds to the same name item in the optimization parameter table * 2 Corresponds to the same name item in the optimization engine parameter table * 3 Corresponds to the same name item in the time series optimization parameter

20. Summary The invention described in this article includes, for example: Two or more items described as items of different examples in the following description may be used in combination. Before describing specific examples, an overview of the system configuration will be described first.

  FIG. 251 is a diagram illustrating a device configuration of the power control system 1 according to an embodiment. The power control system 1 includes a server device 10, a terminal device 20, a facility 30, a server device 40, and a server device 50. The power control system 1 relates to a so-called HEMS. In this example, the server apparatus 10 controls a plurality of HEMSs. However, in order to simplify the drawing, only a single HEMS is shown in FIG. 251, that is, only a single house is shown. In each house, a terminal device 20, a facility 30, and a server device 50 are provided.

  The facility 30 is a device that consumes or supplies power. The facility 30 includes at least two of an electric device, a power supply device, a heat supply device, and an EV (electric vehicle). A specific example of the facility 30 may be different in each house. For example, the EV 30 may be included in the facility 30 of a certain house, and the EV may not be included in the facility 30 of another house. The electrical equipment is, for example, a so-called home appliance. The power supply device is a device that supplies power to another device, and includes, for example, at least one of PV (solar cell), FC (fuel cell), and stationary storage battery. The heat supply device is a device for supplying heat (specifically, for example, hot water) to a house, and includes, for example, a cogeneration device using a fuel cell.

  The terminal device 20 provides a UI. That is, information provided from the system (for example, the server device 10) is provided to the user, and an instruction input to the system is received from the user. The terminal device 20 is, for example, a tablet terminal or a smartphone. The terminal device 20 is a terminal device in the above-described embodiment. The terminal device 20 includes a processor for executing processing, a memory and storage for storing data and programs, a communication interface for communicating via a network, and an input / output device for providing a UI.

  The server device 50 controls and manages the facility 30. The server device 50 collects the record data of the control of the facility 30 and the environmental information (for example, temperature and humidity) acquired by the sensor network and provides them to the server device 10. Further, the server device 50 controls the facility 30 according to the control rule provided from the server device 10. The server device 50 is a home server in the above-described embodiment. Note that a single device may have the functions of the terminal device 20 and the server device 50.

  The server device 10 analyzes various data and provides the facility 30 with a control rule (that is, control information) for controlling the facility 30. The data analyzed by the server device 10 includes, for example, performance data (including EV travel data) of the facility 30, environmental information collected by the sensor network, and weather data. The server device 10 is the information management server in the above-described embodiment. The server device 40 is a server device that provides various data. The server device 10, the server device 40, and the server device 50 have a processor that executes processing, a memory and storage that store data and programs, and a communication interface that communicates via a network.

  FIG. 252 is a diagram illustrating a functional configuration of the power control system 1. The power control system 1 includes a database unit 51, a prediction unit 52, a control rule generation unit 53, a device management unit 54, and a UI unit 55. The database unit 51 stores data used for generating a prediction scenario. The database unit 51 stores various data. The database unit 51 includes an EV travel data storage unit 511, a weather data storage unit 512, a calendar storage unit 513, and an analysis DB 514. The EV travel data storage unit 511 is data relating to EV travel, and indicates, for example, which EV traveled over which distance. The meteorological data storage unit 512 is data related to the weather, and shows, for example, past weather information and future forecasts. The calendar storage unit 513 is data relating to a calendar, and indicates, for example, a date, a day of the week, and a holiday. The analysis DB 514 is a database that collects data used for generating a prediction scenario. The analysis DB 514 includes data acquired from the EV travel data storage unit 511, data acquired from the weather data storage unit 512, data input from the UI unit 55, and data collected by the device management unit 54.

  The prediction unit 52 generates a prediction scenario. The prediction unit 52 includes a PV power generation prediction unit 521, a power demand prediction unit 522, an EV travel prediction unit 523, a heat demand prediction unit 524, and a prediction scenario storage unit 525. The PV power generation prediction unit 521 generates a PV power generation prediction scenario. The power demand prediction unit 522 generates a power demand prediction scenario by the electric device. The EV travel prediction unit 523 generates an EV travel prediction scenario. The heat demand prediction unit 524 generates a heat demand prediction scenario. The prediction scenario storage unit 525 stores the prediction scenarios generated by the PV power generation prediction unit 521, the power demand prediction unit 522, the EV travel prediction unit 523, and the heat demand prediction unit 524.

  The control rule generation unit 53 generates a control rule using the prediction scenario generated by the prediction unit 52. The control rule generation unit 53 includes an optimization plan calculation unit 531, an optimization plan storage unit 532, a control rule generation unit 533, a control rule storage unit 534, a charge table 535, a CO2 emission management table 536, And an objective function definition table 537. The optimization plan calculation unit 531 optimizes the control plan of the facility 30 using the prediction scenario and various data (for example, the charge table 535, the CO2 emission management table 536, and the objective function definition table 537). An optimized control plan is called an optimization plan. The optimization plan storage unit 532 stores the optimization plan. The control rule generation unit 533 generates a control rule based on the optimization plan. The control rule storage unit 534 stores the control rule generated by the control rule generation unit 533.

  The device management unit 54 controls and manages the facility 30. The device management unit 54 includes a state collection unit 541, a life log storage unit 542, a prediction / result determination unit 543, a control rule execution management unit 544, and a device control unit 545. The state collection unit 541 collects information indicating various states (for example, power consumption of home appliances, PV power generation, room temperature, etc., that is, actual values) from each of the facilities 30. The life log storage unit 542 stores information collected by the state collection unit 541. The prediction / result determination unit 543 determines which of the prediction scenario and the result value is used. The control rule execution management unit 544 manages the execution of the control rule according to the control rule generated by the control rule generation unit 53, the prediction scenario, and the actual value. The device control unit 545 controls each of the facilities 30 according to the control rule.

  The UI unit 55 provides a UI. The UI unit 55 includes a control plan display unit 551, a mode selection unit 552, and a device reservation unit 553. The control plan display unit 551 displays the control plan. The control plan is generated from a control rule, for example. The mode selection unit 552 selects one optimization mode among the plurality of optimization modes according to a user instruction. The device reservation unit 553 makes a reservation for the facility 30 (for example, reservation for use of EV, reservation for use of hot water, stop reservation for a specific electric device, etc.).

  A specific example of the PV power generation prediction unit 521 is the PV supply scenario creation submodule (section 6.2.1) in the above-described embodiment. A specific example of the power demand prediction unit 522 is a home appliance power demand scenario creation submodule (Section 6.1.1). A specific example of the EV travel prediction unit 523 is an EV power demand scenario creation submodule (Section 6.1.2). A specific example of the heat demand prediction unit 524 is a heat demand scenario creation submodule (Section 6.1.3). A specific example of the optimization plan calculation unit 531 is an optimization plan submodule (section 6.4.2). A specific example of the control rule generation unit 533 is a control rule generation submodule (section 6.4.3).

  In one embodiment, the database unit 51 is implemented in the server device 40. The prediction unit 52 and the control rule generation unit 53 are mounted on the server device 10. The device management unit 54 and the UI unit 55 are mounted on the terminal device 20. In each device, the functions shown in FIG. 252 are executed by a program.

20.1. Example 1
Example 1 relates to a technique for generating a prediction scenario using a statistical method. Specifically, Example 1 is described as follows.

(Example 1-1)
Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. A first calculation means for calculating a plurality of prediction scenarios and a probability of occurrence of each of the plurality of prediction scenarios using statistical processing of past actual values in each of the facilities;
Second calculation means for calculating a probability that a combination of specific prediction scenarios simultaneously occurs in the facility using the occurrence probability calculated by the first calculation means;
An information processing system comprising: an output unit that outputs a combination of prediction scenarios with the maximum occurrence probability calculated by the second calculation unit as an integrated scenario.

  A device serving as a power load is a device that consumes power. A device to be controlled is a device in which driving such as on and off is controlled by the information processing system. Devices serving as power loads are, for example, home appliances, LIBs, and EVs. Devices to be controlled are, for example, LIB, EV, FC, controllable home appliances (for example, air conditioners, models that can be controlled by EchonetLite), for example, LIB is a device that becomes a power load, and It is also a device to be controlled.

  Specific examples of the first calculating means include a submodule that generates a prediction scenario described in section 5.1.1. Specifically, a home electric power demand scenario creation submodule, an EV power demand scenario creation submodule, and heat This is a demand scenario creation submodule (ie, the power demand prediction unit 522 in FIG. 252). The outline of the home appliance power demand scenario creation submodule is described in section 6.1.1 (for example, FIG. 36), and in section 13.2.1, the demand forecast function for home appliance power (prediction scenario creation) Is described in detail. The EV power demand scenario creation submodule is described in sections 6.1.2 and 13.2.2.

  A specific example of the second calculation means and the output means is the integrated scenario creation submodule described in section 5.4.1. The integrated scenario creation submodule is described in sections 6.4.1 and 13.4.1.

  A specific example of the information processing system is the server apparatus 10 in FIG. 252 (information management server in FIG. 3). Note that this information processing system may be physically configured by a single device, or a plurality of device groups (for example, a server installed in each house and an information management server on the network) connected via a network. ). This information processing system may include a server device on a network (may include a server device on a so-called cloud), and only at least one of a server device and a terminal device installed in each house. It may be constituted by.

  Because users take indeterminate behavior, using a single prediction scenario can cause mispredictions. However, in this example, the accuracy of prediction can be increased by using a plurality of prediction scenarios each having an occurrence probability.

(Example 1-2)
A first acquisition that divides past actual values in each of the facilities into a plurality of categories according to a predetermined criterion, and acquires a supply and demand pattern for each of the plurality of categories from the actual values corresponding to each of the plurality of categories Having means,
The first calculation unit generates the plurality of forecast scenarios by using a pattern corresponding to a period to be predicted among the supply and demand patterns acquired by the first acquisition unit. The information processing system according to 1-1.

  Section 6.1.1.1.1 explains that past performance values are divided into a plurality of categories and that a basic demand pattern corresponding to the period to be forecasted is used ( FIG. 35). Note that the unit of the period to be predicted is not limited to one day. The prediction may be performed in units other than days, such as weeks, hours, and minutes.

(Example 1-3)
Second acquisition means for acquiring a prediction of the environment for the period to be predicted;
Third acquisition means for acquiring environmental information of a category corresponding to the period to be predicted from a database in which environmental information indicating the environment when the actual value is obtained is stored in the plurality of categories. When,
And third calculating means for calculating a difference between the environmental information acquired by the second means and the environmental information acquired by the third acquiring means,
The information processing system according to Example 1-2, wherein the first calculation unit generates the plurality of prediction scenarios by using the difference calculated by the pattern and the third calculation unit.

  Specific examples of the second acquisition unit, the third acquisition unit, and the third calculation unit are home appliance power demand scenario creation submodules. For example, in FIG. 36, a weather forecast is obtained from a weather information company, a future basic demand pattern is set in accordance with the obtained weather forecast, and a forecast scenario using this and a temperature difference (an example of a difference in environmental information). Is described.

(Example 1-4)
The first calculation means assumes a predetermined probability distribution for the difference, divides the difference into a plurality of sections in the probability distribution, and uses an occurrence probability corresponding to each representative value of the plurality of sections, The information processing system according to Example 1-3, wherein occurrence probability of each of the plurality of prediction scenarios is calculated.

  Assuming a predetermined probability distribution (for example, a normal distribution) for the difference in environmental information, dividing the difference into a plurality of sections in this probability distribution, and using the occurrence probability corresponding to the representative value of each section is 6.1. Sections 1.4.2 and 6.1.1.14.3 (FIG. 43).

(Example 1-5)
The information processing system according to Example 1-4, wherein the first calculation unit generates the plurality of prediction scenarios by adding or subtracting representative values of the plurality of sections to the pattern.

  Adding / subtracting the representative value of each section to the basic demand pattern is described in section 6.1.1.1.4.1.

(Example 1-6)
Examples 1-1 to 5 having optimization means for optimizing a control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle using the integrated scenario output by the output means The information processing system according to any one of the above.

  A specific example of the optimization means is the optimization plan submodule described in section 5.4.1 (that is, the optimization plan calculation unit 531 in FIG. 252).

(Example 1-7)
Generating means for generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the control plan optimized by the optimization means The information processing system according to Example 1-6.

  A specific example of the generation means is the control rule generation submodule described in section 5.4.1 (that is, the control rule generation unit 533 in FIG. 252).

(Example 1-8)
On the computer,
Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. Calculating the occurrence probability of each of the plurality of prediction scenarios and the plurality of prediction scenarios using statistical processing of past actual values in each of the facilities;
Calculating a probability that a combination of specific prediction scenarios simultaneously occurs in the plurality of models using the calculated occurrence probability; and
Outputting the combination of the predicted scenarios with the maximum occurrence probability as an integrated scenario.

(Example 1-9)
Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. Calculating the occurrence probability of each of the plurality of prediction scenarios and the plurality of prediction scenarios using statistical processing of past actual values in each of the facilities;
Calculating a probability that a combination of specific prediction scenarios will occur simultaneously in the facility using the calculated occurrence probability;
Outputting the combination of the predicted scenarios with the maximum occurrence probability as an integrated scenario.

(Example 1-10)
The information processing system according to any one of Examples 1-1 to 7,
A communication terminal that communicates with the information processing system;
A power control system comprising the equipment.

20.2. Example 2
Example 2 relates to a technique for generating a prediction scenario reflecting the use schedule of a device. Specifically, Example 2 is described as follows.

(Example 2-1)
Regarding at least one of the facilities in a house having a facility including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles Acquisition means for acquiring the use schedule of
An information processing system comprising: generation means for generating a power supply / demand prediction scenario for each of the facilities, reflecting the use schedule acquired by the acquisition means.

  A specific example of the acquisition unit is the analysis DB 514 in FIG. The analysis DB 514 acquires information indicating device reservation, that is, a use schedule, input from the device reservation unit 553, and stores the acquired information.

  Specific examples of the generation means are an EV power demand prediction scenario creation submodule and a heat demand prediction submodule. Section 6.1.2.3 describes converting EV usage reservations into electricity demand. The heat demand prediction submodule is described in section 6.1.3. However, in the heat demand prediction submodule, the use reservation may be converted into the heat demand amount and reflected in the prediction in the same manner as the EV.

  For example, when there are multiple power supply devices, such as when a house is equipped with both PV and FC, there is a surplus in power generation due to daily use. If the use schedule is input by the user, it is possible to achieve both user convenience and effective use of electric power by generating a prediction scenario (and thus a control rule) reflecting this.

(Example 2-2)
The information processing system according to Example 2-1, wherein the acquisition unit acquires the use schedule according to a user operation input.

  The use schedule may be input according to the operation of the device reservation unit 553 by the user.

(Example 2-3)
The information processing system according to Example 2-1, wherein the acquisition unit acquires the use schedule based on a periodic use pattern extracted from data indicating past use results.

  In this example, the server device 10 uses a periodic usage pattern (for example, EV during the day from Monday to Friday) from data indicating past usage records stored in the life log storage unit 542. However, the usage frequency is low on Saturdays and Sundays. The server device 10 converts the extracted usage pattern into power demand and reflects it in the generation of the prediction scenario.

(Example 2-4)
The generation means uses the use schedule for a period when the use schedule is acquired by the acquisition means, and uses a statistical process of past actual values for periods other than the period when the use schedule is acquired. The information processing system according to any one of Examples 2-1 to 3, wherein a prediction scenario is generated.

  In this example, the server device 10 is extracted from the period when the use schedule is acquired (for example, the period when the use schedule is input from the device reservation unit 553, or the past use record stored in the life log storage unit 542). For a period in which use is indicated by a periodic use pattern), the use schedule is converted into power demand, and for other periods, a prediction scenario is generated by statistical processing of past actual values.

(Example 2-5)
The power supply device includes a solar power generation device,
The information processing system according to any one of Examples 2-1 to 4, wherein the heat supply device includes a cogeneration device.

(Example 2-6)
Using the prediction scenario generated by the generating means, optimizing means for optimizing a control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle;
Generating means for generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the control plan optimized by the optimization means; The information processing system according to any one of Examples 2-1 to 5.

  A specific example of the optimization means is the optimization plan submodule described in section 5.4.1 (that is, the optimization plan calculation unit 531 in FIG. 252). A specific example of the generation means is the control rule generation submodule described in section 5.4.1 (that is, the control rule generation unit 533 in FIG. 252).

(Example 2-7)
On the computer,
Regarding at least one of the facilities in a house having a facility including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles To get a usage plan for
A step of generating a power supply / demand prediction scenario for each of the facilities, reflecting the acquired use schedule.

(Example 2-8)
Regarding at least one of the facilities in a house having a facility including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles To get a usage plan for
Reflecting the acquired use schedule, and generating a power supply / demand prediction scenario for each of the facilities.

(Example 2-9)
The information processing system according to any one of Examples 2-1 to 6,
A communication terminal that communicates with the information processing system;
A power control system comprising the equipment.

20.3. Example 3
Example 3 relates to a technique for generating a control rule using an optimization mode selected from one or more optimization modes including an environmental mode. Specifically, Example 3 is described as follows.

(Example 3-1)
Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. First generation means for generating a prediction scenario of
An optimization unit that optimizes a control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle using the prediction scenario generated by the first generation unit, the optimization A plurality of algorithms corresponding to one or more operation modes including a first mode for minimizing carbon dioxide emission for generating electric power used in the facility, wherein the one or more operation modes Optimization means for performing the optimization according to one operation mode selected from:
Second generation means for generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the control plan optimized by the optimization means And an information processing system.

  Specific examples of the first generation means include a sub-module that generates a prediction scenario described in section 5.1.1. Specifically, a home electric power demand scenario creation sub-module, an EV power demand scenario creation sub-module, and heat This is a demand scenario creation submodule (ie, the power demand prediction unit 522 in FIG. 252). The outline of the home appliance power demand scenario creation submodule is described in section 6.1.1 (for example, FIG. 36), and in section 13.2.1, the demand forecast function for home appliance power (prediction scenario creation) Is described in detail. The EV power demand scenario creation submodule is described in sections 6.1.2 and 13.2.2.

  A specific example of the optimization means is the optimization plan submodule described in section 5.4.1 (that is, the optimization plan calculation unit 531 in FIG. 252). A specific example of the first mode for minimizing the carbon dioxide emission is an environmental (eco) mode, which is described in section 3.2 and FIG.

  A specific example of the second generation means is the control rule generation submodule described in section 5.4.1 (that is, the control rule generation unit 533 in FIG. 252).

  The previous HEMS was aimed at maximizing power sales and improving operating efficiency, but according to this example, it creates new value of reducing carbon dioxide emissions.

(Example 3-2)
For each of the facilities, a first calculation means for calculating a plurality of prediction scenarios of power supply and demand and the occurrence probability of each of the plurality of prediction scenarios using statistical processing of past actual values in each of the facilities;
Second calculation means for calculating a probability that a combination of specific prediction scenarios among the prediction scenarios generated by the generation means in the facility is generated at the same time, using the occurrence probability calculated by the first calculation means;
Output means for outputting, as an integrated scenario, a combination of prediction scenarios with the maximum occurrence probability calculated by the second calculation means;
The information processing system according to Example 3-1, wherein the optimization unit performs the optimization using the integrated scenario output by the output unit.

  Specific examples of the first calculating means include a submodule that generates a prediction scenario described in section 5.1.1. Specifically, a home electric power demand scenario creation submodule, an EV power demand scenario creation submodule, and heat This is a demand scenario creation submodule (ie, the power demand prediction unit 522 in FIG. 252). The outline of the home appliance power demand scenario creation submodule is described in section 6.1.1 (for example, FIG. 36), and in section 13.2.1, the demand forecast function for home appliance power (prediction scenario creation) Is described in detail. The EV power demand scenario creation submodule is described in sections 6.1.2 and 13.2.2.

  A specific example of the second calculation means and the output means is the integrated scenario creation submodule described in section 5.4.1. The integrated scenario creation submodule is described in sections 6.4.1 and 13.4.1.

(Example 3-3)
The algorithm corresponding to the first mode is an algorithm for minimizing a carbon dioxide emission amount calculated based on a carbon dioxide emission amount generated by gas combustion in the electric device, the power supply device, and the heat supply device. The information processing system according to Example 3-1 or 2, wherein:

  Regarding the algorithm for minimizing the carbon dioxide emission, in the objective function described in Section 6.4.2.3.3 and FIG. 78, the items of the grid power unit price and the fuel unit price correspond to the exhausted carbon dioxide amount. It describes that a numerical value is set (FIG. 79).

(Example 3-4)
The power supply device
A reformer to obtain hydrogen gas by steam reforming the fuel;
A fuel cell unit that generates electric power by reacting hydrogen obtained by the reformer with oxygen in the air, and
The information processing according to Example 3-3, wherein the algorithm corresponding to the first mode is an algorithm that minimizes a carbon dioxide emission amount calculated based on a carbon dioxide generation amount in the reformer. system.

  In this example, the server apparatus 10 may consider the carbon dioxide generation amount in the reformer in the cogeneration apparatus. For the carbon dioxide generation amount in the reformer, for example, in the objective function (FIG. 78) indicating the cost of hot water supply with gas and the cost of exhaust heat, a numerical value corresponding to the amount of discharged carbon dioxide may be set in the item of fuel unit price.

(Example 3-5)
The algorithm corresponding to the first mode is an algorithm that minimizes the carbon dioxide emission obtained based on the power composition of the grid power used in the house. Information processing system.

  In this example, the server apparatus 10 may consider the carbon dioxide generation amount resulting from the grid power. The amount of carbon dioxide generated due to the grid power is estimated based on the power composition. The power composition refers to the ratio of power by power generation method. For example, the ratio of thermal power generation, hydroelectric power generation, and nuclear power generation in the grid power is provided from a database or the like. Further, for each power generation method, the amount of carbon dioxide generated per unit power is provided from a database or the like.

(Example 3-6)
On the computer,
Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. Generating a forecast scenario for
A step of optimizing a control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle using the generated prediction scenario, wherein the optimization algorithm includes the facility One operation selected from the one or more operation modes, including a plurality of algorithms corresponding to one or more operation modes including a first mode for minimizing carbon dioxide emission for generating electric power used in Performing the optimization according to a mode;
Generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the optimized control plan. .

(Example 3-7)
Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. Generating a forecast scenario for
A step of optimizing a control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle using the generated prediction scenario, wherein the optimization algorithm includes the facility One operation selected from the one or more operation modes, including a plurality of algorithms corresponding to one or more operation modes including a first mode for minimizing carbon dioxide emission for generating electric power used in Performing the optimization according to a mode;
Generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the optimized control plan.

(Example 3-8)
The information processing system according to any one of Examples 3-1 to 5,
A communication terminal that communicates with the information processing system;
A power control system comprising the equipment.

20.4. Example 4
Example 4 relates to a technique for generating a control rule using an optimization mode selected from a plurality of optimization modes each having a different objective function. Specifically, Example 4 is described as follows.

(Example 4-1)
Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. First generation means for generating a prediction scenario of
Using the prediction scenario generated by the first generation unit, the control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle is optimized by a probability programming method using an objective function. The optimization algorithm includes a plurality of algorithms corresponding to a plurality of operation modes each using a different objective function, and the optimization algorithm is optimized according to one operation mode selected from the plurality of operation modes. An optimization means for performing
Second generation means for generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the control plan optimized by the optimization means And an information processing system.

  Specific examples of the first generation means include a sub-module that generates a prediction scenario described in section 5.1.1. Specifically, a home electric power demand scenario creation sub-module, an EV power demand scenario creation sub-module, and heat This is a demand scenario creation submodule (ie, the power demand prediction unit 522 in FIG. 252). The outline of the home appliance power demand scenario creation submodule is described in section 6.1.1 (for example, FIG. 36), and in section 13.2.1, the demand forecast function for home appliance power (prediction scenario creation) Is described in detail. The EV power demand scenario creation submodule is described in sections 6.1.2 and 13.2.2.

  A specific example of the optimization means is the optimization plan submodule described in section 5.4.1 (that is, the optimization plan calculation unit 531 in FIG. 252). Specific examples of the plurality of operation modes are a plurality of optimization modes, which are described in section 3.2 and FIG.

  A specific example of the second generation means is the control rule generation submodule described in section 5.4.1 (that is, the control rule generation unit 533 in FIG. 252).

  According to this example, more flexible power control according to the user's request is provided.

(Example 4-2)
The plurality of objective functions includes a first objective function indicating a cost related to power purchase,
The plurality of operation modes include a first mode that minimizes carbon dioxide emissions,
In the first mode, a numerical value corresponding to carbon dioxide emission is used as a coefficient in the first objective function. The information processing system according to Example 4-1.

  A specific example of the first objective function is an objective function indicating the cost related to power purchase, which is described in FIG. A specific example of the first mode is the environment mode described in FIG.

(Example 4-3)
The plurality of objective functions include a first objective function indicating a cost related to power purchase and a second objective function indicating a cost related to fuel,
The plurality of operating modes includes a second mode that minimizes power and fuel costs;
In the second mode, a numerical value corresponding to the cost of system power is used as a coefficient in the first objective function, and a numerical value corresponding to the cost of fuel is used as a coefficient in the second objective function. 4-1. The information processing system according to 4-1.

  A specific example of the second objective function is an objective function shown in FIG. 78, which indicates the fuel cost dropped on the fuel cell and the cost related to start / stop. A specific example of the second mode is the economic mode described in FIG.

(Example 4-4)
The plurality of objective functions includes a first objective function indicating a cost related to power purchase,
The plurality of operation modes include a third mode for minimizing power purchase amount,
In the third mode, a numerical value corresponding to the cost of system power is used as a coefficient in the first objective function. The information processing system according to Example 4-1.

  A specific example of the third mode is the self-sufficient mode described in FIG.

(Example 4-5)
The plurality of objective functions includes a third objective function indicating a cost related to switching and remaining amount of the storage battery in the electric vehicle,
The plurality of operation modes include a fourth mode for maximizing the remaining power in the storage battery,
In the fourth mode, a numerical value corresponding to a penalty with respect to a remaining battery level is used as a coefficient in the third objective function. The information processing system according to Example 4-1.

  A specific example of the third objective function is an objective function described in FIG. 78 that indicates the cost related to switching and remaining amount of the EV storage battery. A specific example of the fourth mode is the charge priority mode described in FIG.

(Example 4-6)
The power supply device includes a stationary storage battery,
The plurality of objective functions includes a fourth objective function indicating a cost related to switching and remaining amount of the storage battery,
The plurality of operation modes include a fourth mode for maximizing the remaining power in the storage battery,
In the fourth mode, a numerical value corresponding to a penalty with respect to a remaining battery level is used as a coefficient in the fourth objective function. The information processing system according to Example 4-1.

  A specific example of the fourth objective function is an objective function shown in FIG. 78 that indicates the cost related to switching and remaining amount of the stationary storage battery.

(Example 4-7)
The plurality of objective functions indicate a first objective function indicating a cost related to power purchase, and a third objective function indicating a cost related to switching and remaining capacity of a storage battery in the electric vehicle, or a cost related to switching and remaining capacity of a stationary storage battery. Including a fourth objective function,
The plurality of operation modes include a fifth mode in which the grid power is disconnected and the autonomous operation period is maximized,
In the fifth mode, a numerical value that makes the tidal flow from the system power supply zero is used as a coefficient in the first objective function, and a numerical value corresponding to a penalty for the remaining battery capacity of the storage battery or the stationary storage battery in the electric vehicle is The information processing system according to Example 4-1, wherein the information processing system is used.

  A specific example of the fifth mode is the emergency mode described in FIG.

(Example 4-8)
On the computer,
Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. Generating a forecast scenario for
A step of optimizing a control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle using the generated prediction scenario by a probability programming method using an objective function, The optimization algorithm includes a plurality of algorithms corresponding to a plurality of operation modes each using a different objective function, and performing the optimization according to one operation mode selected from the plurality of operation modes;
Generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the optimized control plan. .

(Example 4-9)
Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. Generating a forecast scenario for
A step of optimizing a control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle using the generated prediction scenario by a probability programming method using an objective function, The optimization algorithm includes a plurality of algorithms corresponding to a plurality of operation modes each using a different objective function, and performing the optimization according to one operation mode selected from the plurality of operation modes;
Generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the optimized control plan.

(Example 4-10)
The information processing system according to any one of Examples 4-1 to 7,
A communication terminal that communicates with the information processing system;
A power control system comprising the equipment.

20.5. Example 5
Example 5 relates to a technique for generating a prediction scenario using a statistical method. Specifically, Example 1 is described as follows.

  Examples 5-1 to 5-10 relate to generation of a prediction scenario. Specifically, it relates to the function of the home appliance power demand scenario creation submodule whose outline is explained in section 6.1.1.

(Example 5-1)
First generation means for generating a plurality of prediction scenarios of power supply and demand in a target period to be predicted for each of the facilities in a house having facilities including a load device and a control target device;
First calculation means for calculating the probability of occurrence of the prediction scenario for each of the plurality of prediction scenarios;
Selection means for selecting one prediction scenario for each of the facilities based on the occurrence probability calculated by the first calculation means;
An information processing system comprising: a second generation unit configured to generate a control plan for the facility based on a combination of prediction scenarios selected by the selection unit.

(Example 5-2)
The first generation unit generates the prediction scenario for at least one device of the facilities by using a statistical processing result of a past actual power demand value in the device. 1. The information processing system according to 1.

(Example 5-3)
The first generation unit classifies the actual value into a plurality of patterns based on a predetermined reference for at least one device of the equipment, and uses a pattern corresponding to the target period among the plurality of patterns. The information processing system according to Example 5-2, wherein the prediction scenario is generated.

(Example 5-4)
The information processing system according to Example 5-3, wherein the standard includes matters related to a calendar.

(Example 5-5)
The information processing system according to Example 5-3 or 4, wherein the standard includes a matter related to climate.

(Example 5-6)
The first generation means includes, for at least one device among the facilities, a statistical processing result of a difference between a past prediction scenario of a power demand in the device and an actual value corresponding to the prediction scenario, and a combination of the patterns The information processing system according to any of Examples 5-3 to 5, wherein the prediction scenario is generated by:

(Example 5-7)
The first generation means determines a parameter in a predetermined mathematical model for the difference using a statistical processing result of the difference, and predicts in the pattern using the mathematical model using the determined parameter. The information processing system according to Example 5-6, wherein a fluctuation range of the value is calculated, and the prediction scenario is generated using the calculated fluctuation range.

(Example 5-8)
The first calculation means assumes a predetermined probability distribution for the difference, divides the difference into a plurality of sections in the probability distribution, and uses an occurrence probability corresponding to each representative value of the plurality of sections, The information processing system according to Example 5-7, wherein the occurrence probability of each of the plurality of prediction scenarios is calculated.

(Example 5-9)
The information processing system according to Example 5-8, wherein the first generation unit generates the plurality of prediction scenarios by adjusting the representative value with respect to the pattern.

(Example 5-10)
The first generation means uses, for at least one device among the facilities, a statistical processing result of a difference between the prediction scenario and the actual value corresponding to a past prediction scenario of power demand in the device. The information processing system according to Example 5-2, wherein the prediction scenario is generated.

  Examples 5-11 to 5-14 relate to generation of a prediction scenario. Specifically, it relates to the function of the EV power demand scenario creation submodule outlined in section 6.1.2.

(Example 5-11)
The facility includes an electric vehicle,
Obtaining means for obtaining the use schedule of the electric vehicle;
The said 1st production | generation means produces | generates the said prediction scenario using the use plan acquired by the said acquisition means about the said electric vehicle. Any one of Examples 5-1 thru | or 10 characterized by the above-mentioned. Information processing system.

(Example 5-12)
The information processing system according to Example 5-11, wherein the acquisition unit acquires the use schedule according to a user operation input.

(Example 5-13)
From the data indicating the past use record of the electric vehicle, it has an extraction means for extracting a periodic use pattern,
The information processing system according to Example 5-11, wherein the acquisition unit acquires the use pattern extracted by the extraction unit as the use schedule.

(Example 5-14)
The first generation means extracts a use record that satisfies a predetermined condition from the use record of the day corresponding to the target period from the past use record of the electric vehicle divided by day of the week. The information processing system according to any one of Examples 5-11 to 13, wherein the prediction scenario is generated based on the actual usage record.

  Examples 5-15 to 5-19 relate to generation of a prediction scenario. Specifically, the function of the EV power demand scenario creation sub-module outlined in Section 6.1.2 is applied to the hot water supply device.

(Example 5-15)
The facility includes a hot water supply device,
Obtaining means for obtaining a use schedule of the hot water supply device;
The said 1st production | generation means produces | generates the said prediction scenario using the use plan acquired by the said acquisition means about the said hot water supply apparatus. Any one of the examples 5-1 thru | or 10 characterized by the above-mentioned. Information processing system.

(Example 5-16)
The information processing system according to Example 5-15, wherein the acquisition unit acquires the use schedule according to a user operation input.

(Example 5-17)
From the data indicating the past usage record of the hot water supply device, it has an extraction means for extracting a periodic use pattern,
The information processing system according to Example 5-15, wherein the acquisition unit acquires the use pattern extracted by the extraction unit as the use schedule.

(Example 5-18)
The first generation unit extracts a use record that satisfies a predetermined condition from the use record of the day corresponding to the target period from the past use record of the hot water supply apparatus divided for each day of the week, and the extracted The information processing system according to any one of Examples 5-15 to 17, wherein the prediction scenario is generated based on the actual usage record.

(Example 5-19)
The first calculation means assigns a predetermined occurrence probability to a prediction scenario corresponding to the use schedule, and assigns the remaining occurrence probability evenly to the prediction scenario generated based on the usage record, thereby The information processing system according to Example 5-14 or 18, wherein an occurrence probability is calculated for each of the prediction scenarios.

  Examples 5-20 to 5-28 relate to the generation of a prediction scenario. Specifically, it relates to the function of the heat demand scenario creation submodule outlined in section 6.1.3.

(Example 5-20)
The facility includes a heat supply device,
The first generation means generates, for the heat supply device, a plurality of prediction scenarios of heat supply and demand in the target period using a statistical processing result of a past heat demand actual value in the device. The information processing system according to any one of Examples 5-1 to 19.

(Example 5-21)
The first generation unit classifies the actual values for the heat supply device into a plurality of categories based on a predetermined criterion, and uses the actual values belonging to the category corresponding to the target period among the plurality of categories. The information processing system according to Example 5-20, wherein the prediction scenario is generated.

(Example 5-22)
The information processing system according to Example 5-21, wherein the reference includes a matter related to a calendar.

(Example 5-23)
The information processing system according to Example 5-21 or 22, wherein the standard includes matters related to climate.

(Example 5-24)
The first generation unit uses the actual value that satisfies a predetermined condition that a difference between the target period and the maximum temperature among the actual values belonging to the category corresponding to the target period for the heat supply device, and the prediction scenario. The information processing system according to any of Examples 5-21 to 23, wherein:

(Example 5-25)
The first generation means generates the prediction scenario by using a statistical processing result of a difference between the prediction scenario and the actual value corresponding to the prediction scenario of the heat demand in the heat supply device. The information processing system according to Example 5-20, characterized by:

(Example 5-26)
The first generation means assumes a predetermined probability distribution with respect to the difference, divides the difference into a plurality of sections in the probability distribution, and uses the representative values of the plurality of sections to use the plurality of prediction scenarios. The information processing system according to Example 5-25, wherein the information processing system is generated.

(Example 5-27)
The information processing system according to Example 5-25 or 26, wherein the first calculation means calculates the occurrence probability using the statistical processing result.

(Example 5-28)
The information processing system according to Example 5-26, wherein the first calculation unit calculates the occurrence probability using an occurrence probability corresponding to a representative value of each of the plurality of sections in the probability distribution. .

  Examples 5-29 to 5-30 relate to selection of combinations of prediction scenarios. Specifically, it relates to the function of the integrated scenario creation submodule whose outline is explained in section 6.4.1.

(Example 5-29)
A second calculating means for calculating a probability that a combination of specific prediction scenarios will occur at the same time in the facility using the occurrence probability calculated by the first calculating means;
The information processing system according to any one of Examples 5-1 to 28, wherein the selection unit selects a combination of prediction scenarios having the maximum occurrence probability calculated by the second calculation unit.

(Example 5-30)
From the time series data of the error corresponding to the prediction scenario obtained for each of the equipment, having an estimation means for estimating the correlation between each of the equipment,
The information processing system according to Example 5-29, wherein the second calculating unit calculates the occurrence probability of the combination from the correlation and the multidimensional normal density function obtained by the estimating unit.

  Example 5-31 to Example 5-34 relate to timing (timing or frequency of various calculations) for performing estimation using a model. This point will not be described in detail in the above embodiment, and will be described below. In the system in the embodiment, what kind of processing is performed when a prediction scenario (prediction value) is out of place, that is, when a difference between the prediction scenario and the actual value exceeds a predetermined range (threshold value), for example. It becomes a problem.

  For example, there are the following three methods for dealing with this problem. The first method is a method in which the parameters of the model are updated frequently (for example, at a time interval corresponding to a mesh), and a prediction that always reflects the latest actual value is performed. This method has an advantage that prediction can be performed with higher accuracy, but has a problem that the calculation load increases. The second method does not update the model parameters as long as the difference between the forecast scenario and the actual value is within a predetermined range, and if the difference between the forecast scenario and the actual value exceeds this range, It is a method to update parameters. In the third method, the second method temporarily stops the control when the difference between the forecast scenario and the actual value exceeds this range, and performs the control when the actual value reenters the planned value. It is a way to resume.

(Example 5-31)
The first generation means includes
For at least one device of the equipment, generate the prediction scenario using a statistical processing result of a difference between a past prediction scenario of power demand in the device and a performance value corresponding to the prediction scenario,
Determining a parameter in a predetermined mathematical model for the difference using a statistical processing result of the difference;
Using the mathematical model using the determined parameter, calculate the amplitude of the predicted value in the pattern,
The information processing system according to any one of Examples 5-1 to 30, wherein the prediction scenario is generated using the calculated fluctuation width.

(Example 5-32)
The information processing system according to Example 5-31, wherein the first generation unit determines the parameter each time generation of the prediction scenario is instructed.

(Example 5-33)
When the difference between the actual value corresponding to the prediction scenario and the prediction scenario exceeds a predetermined range, the selection unit reselects the prediction scenario based on the actual value,
The information processing system according to Example 5-31, wherein the first calculation unit calculates an occurrence probability of the prediction scenario for each of the reselected prediction scenarios.

(Example 5-34)
When the difference between the actual value corresponding to the prediction scenario and the prediction scenario exceeds a predetermined range, the second generation means generates a control plan for temporarily interrupting the control of the facility,
When the difference between the actual value corresponding to the prediction scenario and the prediction scenario falls within a predetermined range, the second generation means generates a control plan for temporarily interrupting the control of the facility. The information processing system according to Example 5-31.

  Examples 5-35 through 5-38 relate to other embodiments of Examples 5-1 through 5-34. The functions of the submodules of the demand module described in the embodiment may be interchanged or combined with each other. For example, in the home appliance power demand scenario creation submodule, a prediction scenario based on the use schedule may be generated, similarly to the EV demand scenario creation submodule. In addition, in the EV demand scenario creation submodule, a prediction scenario may be generated based on a mathematical model as in the home appliance power demand scenario creation submodule. In addition, the parameters of the mathematical model may be determined by statistical processing of past actual values.

(Example 5-35)
The information processing system according to Example 5-1, wherein the facility includes at least one of a heat supply device and an electric vehicle.

(Example 5-36)
On the computer,
Generating a plurality of prediction scenarios of power supply and demand in a target period to be predicted for each of the facilities in a house having facilities including a load device and a control target device;
For each of the plurality of prediction scenarios, calculating an occurrence probability of the prediction scenario;
Selecting one prediction scenario for each of the facilities based on the calculated probability of occurrence;
Generating a control plan for the facility based on the combination of the selected prediction scenarios.

(Example 5-37)
Generating a plurality of prediction scenarios of power supply and demand in a target period to be predicted for each of the facilities in a house having facilities including a load device and a control target device;
For each of the plurality of prediction scenarios, calculating an occurrence probability of the prediction scenario;
Selecting one prediction scenario for each of the facilities based on the calculated probability of occurrence;
Generating a control plan for the facility based on the combination of the selected prediction scenarios.

(Example 5-38)
The information processing system according to any one of Examples 5-1 to 37;
A communication terminal that communicates with the information processing system;
A power control system comprising the equipment.

DESCRIPTION OF SYMBOLS 1 ... Power control system, 10 ... Server apparatus, 20 ... Terminal apparatus, 30 ... Apparatus, 40 ... Server apparatus, 51 ... Database part, 52 ... Prediction part, 53 ... Control rule production | generation part, 54 ... Equipment management part, 55 ... UI unit, 511 ... EV travel data storage unit, 512 ... weather data storage unit, 513 ... calendar storage unit, 514 ... DB for analysis, 521 ... PV power generation prediction unit, 522 ... power demand prediction unit, 523 ... EV travel prediction unit 524 ... Thermal demand prediction unit, 525 ... Prediction scenario storage unit, 531 ... Optimization plan calculation unit, 532 ... Optimization plan storage unit, 533 ... Control rule generation unit, 534 ... Control rule storage unit, 535 ... Toll table 536: CO2 emission management table, 537 ... objective function definition table, 541 ... state collection unit, 542 ... life log storage unit, 543 ... prediction / result determination unit, 544 ... control level Le execution management unit, 545 ... device control unit, 551 ... control plan display unit, 552 ... mode selection unit, 553 ... equipment reservation unit

Claims (5)

Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. A first calculation means for calculating a plurality of prediction scenarios and a probability of occurrence of each of the plurality of prediction scenarios using statistical processing of past actual values in each of the facilities;
Second calculation means for calculating a probability that a combination of specific prediction scenarios simultaneously occurs in the facility using the occurrence probability calculated by the first calculation means;
An information processing system comprising: an output unit that outputs a combination of prediction scenarios with the maximum occurrence probability calculated by the second calculation unit as an integrated scenario. Regarding at least one of the facilities in a house having a facility including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles Acquisition means for acquiring the use schedule of
An information processing system comprising: generation means for generating a power supply / demand prediction scenario for each of the facilities, reflecting the use schedule acquired by the acquisition means. Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. Generating means for generating a prediction scenario of
An optimization unit that optimizes a control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle using the prediction scenario generated by the generation unit, The algorithm includes a plurality of algorithms corresponding to one or more operation modes including a first mode for minimizing carbon dioxide emissions for generating power used by the facility, and selected from the one or more operation modes Optimization means for performing the optimization according to the one operation mode performed;
Generating means for generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the control plan optimized by the optimization means; Information processing system. Supply and demand of electric power for each of the facilities in a house having facilities including one or more devices serving as power loads and one or more devices to be controlled among electric devices, power supply devices, heat supply devices, and electric vehicles. Generating means for generating a prediction scenario of
Optimizing the control plan of the electric device, the power supply device, the heat supply device, and the electric vehicle by using the prediction scenario generated by the generation means by a probability programming method using an objective function The optimization algorithm includes a plurality of algorithms corresponding to a plurality of operation modes each using a different objective function, and the optimization is performed according to one operation mode selected from the plurality of operation modes. Optimization means to perform,
Generating means for generating a control rule for controlling at least one of the electric device, the power supply device, the heat supply device, and the electric vehicle based on the control plan optimized by the optimization means; Information processing system. First generation means for generating a plurality of prediction scenarios of power supply and demand in a target period to be predicted for each of the facilities in a house having facilities including a load device and a control target device;
First calculation means for calculating the probability of occurrence of the prediction scenario for each of the plurality of prediction scenarios;
Selection means for selecting one prediction scenario for each of the facilities based on the occurrence probability calculated by the first calculation means;
An information processing system comprising: a second generation unit configured to generate a control plan for the facility based on a combination of prediction scenarios selected by the selection unit.