JP2024068782A - Power management system and charge/discharge plan creation method - Google Patents

Abstract

[Problem] To create an optimal charge/discharge plan that is highly likely to satisfy a given constraint even if the prediction is wrong and that is as close as possible to the objective. [Solution] The system includes components 1B-1D that predict power generation and calculate another power generation prediction value taking into account prediction errors, components 2B-2D that predict power demand and calculate another demand prediction value taking into account prediction errors, and a charge/discharge plan creation unit 5 that creates a charge/discharge plan for a storage battery by performing optimization calculations using information on time-of-day electricity rate systems to reduce the electricity procurement fee for purchasing electricity while the multiple power generation prediction values and multiple demand prediction values predicted and calculated as described above and the state of the storage battery satisfy the constraint conditions, and by performing optimization calculations including not only the actual prediction values but also the different prediction values taking into account errors, it is possible to create a charge/discharge plan that is highly likely to satisfy the constraint conditions even if the actual prediction is wrong. [Selected Figure] Figure 2

Description

The present invention relates to a power management system and a method for creating a charge/discharge plan, and is particularly suitable for use in a system that creates a charge/discharge plan for a storage battery that is optimized to reduce power procurement costs.

In recent years, energy management systems that utilize power generation and storage have been developed and widely offered from the perspective of expanding the introduction of renewable energy and adjusting the balance of power supply and demand toward a carbon-free society. For example, a system is known that aims to reduce electricity procurement costs by charging surplus electricity from solar power generation into a storage battery and using it at other times, such as at night. Attention is also being paid to the use of storage batteries installed in electric vehicles (EVs), which are likely to become more widespread in the future.

In this regard, a system has been known that performs power management based on the predicted power generation by a power generation device, the predicted power consumption due to the power load, the usage plan or predicted usage state of an EV equipped with a storage battery, and electricity rates by time period to minimize electricity charges without compromising the convenience of the EV as a form of mobility (see, for example, Patent Documents 1 and 2). In the following description, an EV is referred to as a vehicle, and a storage battery installed in the vehicle is referred to as an on-board storage battery.

The power management system described in Patent Document 1 includes a natural energy power generation device, a power load, a power management device, and a vehicle equipped with a storage battery (vehicle storage battery), and the power management device calculates the amount of power stored in the vehicle storage battery for each hour based on an off-site use schedule including the scheduled time to start off-site use of the vehicle storage battery, the scheduled time to end said use, and a predicted amount of power required for said use, a predicted power consumption of the power load, and a predicted power generation of the natural energy power generation device. It then plans charging of the vehicle storage battery using externally supplied power during a time period when electricity rates are relatively cheap so that the maximum value of the obtained amount of power stored for each hour is a predetermined maximum amount of power stored, and plans charging and discharging of the vehicle storage battery so that the amount of power stored in the vehicle storage battery at the scheduled end time of off-site use is equal to or greater than a predetermined minimum amount of power stored.

In the power supply system described in Patent Document 2, a charge/discharge schedule is calculated based on consumption prediction data, power generation prediction data, and data on the period during which the on-board storage battery is electrically connected to the wiring, calculated based on the vehicle usage history, to indicate the progress of charge and discharge of the power storage means including the on-board storage battery during the prediction period, and the charge/discharge power of the power storage means during the prediction period is controlled according to the calculated charge/discharge schedule. The charge/discharge schedule is calculated by formulating it into a mixed integer programming problem so that the evaluation index (power purchase fee according to the amount of power supplied from the AC power line) used to determine the charge/discharge power of the power storage means becomes a predetermined value.

There are also known systems that aim to optimize the charge and discharge plan of a storage battery from the standpoint of both reducing electricity charges and reducing deterioration of the storage battery (see, for example, Patent Documents 3 to 5). The charge and discharge control device described in Patent Document 3 generates constraint conditions for calculating a charge and discharge plan based on a demand forecast for electricity supplied from a power system in which electricity charges are generated according to use and a storage battery that is charged with electricity from the power system, and generates an objective function that minimizes electricity charges and deterioration of the storage battery to solve an optimization problem, thereby generating a charge and discharge plan for the storage battery that increases the amount of savings in electricity charges and keeps the deterioration of the storage battery low.

In the charge and discharge plan setting described in Patent Document 4, an optimization problem is solved to reduce electricity charges and suppress deterioration of the storage battery based on power demand forecasts, etc., to create an optimal charge and discharge plan for the electricity charges and deterioration level. Patent Document 4 describes solving an optimization problem to find a charge and discharge plan for each specified time period, with the constraints being that the storage battery does not flow reversely to the grid, that the storage battery does not exceed its storage capacity limit, and that the storage battery does not exceed its hourly charge and discharge limit, and with a function for finding an optimal solution that minimizes the sum of the electricity charges and the deterioration level of the storage battery as the objective function.

Patent Document 5 discloses an electric power load leveling device that can optimize the timing of peak power consumption and charging and discharging of a storage battery, and can more efficiently perform load leveling, including power peak shaving, using a storage battery. In the electric power load leveling device described in Patent Document 5, when the predicted required discharge amount (power amount required for peak shaving) is lower than a predetermined value (for example, about 80% SOC), the lithium ion battery is charged using nighttime power. Also, when the predicted required discharge amount is higher than the predetermined value, the lithium ion battery is charged so that charging is completed just before the predicted power peak. This achieves both suppressing deterioration of the lithium ion battery to improve its lifespan, and charging the lithium ion battery using a nighttime rate at which electricity rates are relatively low.

In the systems described in Patent Documents 1 to 5, optimization calculations are performed based on prediction results such as power generation, power demand, and the usage state of the on-board storage battery, and a charge/discharge plan for the storage battery that satisfies given constraints is created. In other words, the charge/discharge plan is created by applying the prediction results to the constraint equations assuming that they are correct. However, the actual power generation state, demand state, and usage state of the on-board storage battery do not always turn out as predicted. Therefore, there is a problem in that when charge/discharge control of the storage battery is performed according to a charge/discharge plan created based on uncertain prediction results, the charge/discharge control may not actually satisfy the constraints.

Robust optimization is known as a modeling technique for optimization problems that can return reliable results even when the data defining the problem is inaccurate or uncertain. Conventionally, systems that apply robust optimization to create power operation plans are also known (see, for example, Patent Documents 6 and 7). However, the techniques described in Patent Documents 6 and 7 do not optimize the charging and discharging plans for storage batteries.

The operation plan creation device described in Patent Document 6 generates forecast data indicating a statistically possible range based on data indicating actual power demand and data indicating actual power generation by renewable energy. Then, a generator operation plan is created by solving a problem that satisfies a constraint equation including the range indicated by the forecast data and predetermined constraints related to the generator specifications and the state of the power system, and brings the value of an objective function, which has at least the cost of suppressing renewable energy, closer to a desired value. Here, problem data is generated by applying input values to the objective function and constraint equation, and the problem data is transformed to apply to a robust optimization method, and a generator operation plan is created based on the robust optimization method for the transformed equation.

In the information processing device described in Patent Document 7, the amount of power procured from each of multiple power procurement means is used as an operational variable, a value including the total cost required to procure an amount of power corresponding to the predicted demand using the multiple power procurement means is used as an objective function, an optimization model having constraints on the amount of power that can be procured from the power procurement means is generated, and a power procurement plan indicating the amount of power to be procured from each of the multiple power procurement means is determined by solving the optimization model using a mathematical optimization method. Here, even if an uncertain constraint is given to the power demand, a robust optimization problem assuming replanning is formulated, and an optimal solution to the formulated robust optimization problem is obtained, thereby determining an optimal power generation plan for the input power demand.

Patent No. 6053554 Patent No. 5672186 JP 2016-73113 A JP 2018-23188 A JP 2016-116401 A JP 2021-33625 A JP 2021-157724 A

The present invention has been made to solve the problems described above, and aims to make it possible to create an optimal charging and discharging plan for a storage battery based at least on power generation and demand forecasts, which is likely to satisfy given constraints even if the forecasts are incorrect, and which comes as close as possible to the objective.

In order to solve the above-mentioned problems, in the present invention, a predicted value of the amount of power generation at each time during a prediction period is obtained, and another predicted value of power generation at each time during the prediction period is calculated using the obtained predicted value of power generation and a prediction error of the amount of power generation expected at each time during the prediction period calculated based on the actual value of the amount of power generation in the past. In addition, a predicted value of the amount of power demand at each time during the prediction period is obtained, and another predicted demand value at each time during the prediction period is calculated using the obtained predicted demand value and a prediction error of the amount of power demand expected at each time during the prediction period calculated based on the actual value of the amount of power demand in the past. Then, using information on the electricity rate system for each time period during the prediction period, an optimization calculation is performed to reduce the power procurement fee for purchasing electricity while the multiple predicted power generation values and multiple predicted demand values obtained and calculated as described above and the state of the storage battery satisfy the constraint conditions set for the predicted power generation value, the predicted demand value, and the state of the storage battery, thereby creating a charging and discharging plan for the storage battery.

According to the present invention configured as described above, optimization calculations are performed using multiple power generation forecast values and multiple demand forecast values, including not only the actual power generation forecast values and demand forecast values acquired initially, but also other power generation forecast values and demand forecast values calculated based on the acquired actual power generation forecast values and demand forecast values. In other words, optimization calculations are performed such that the multiple power generation forecast values and multiple demand forecast values and the state of the storage battery satisfy the constraint conditions and reduce the power procurement fee for purchasing power. As a result, even if the actual power generation forecast and demand forecast are incorrect, it is highly likely that the constraint conditions will be satisfied, and an optimal charging and discharging plan can be created that keeps the power procurement fee for purchasing power as low as possible.

In particular, in the present invention, since the actual power generation forecast value and demand forecast value and the forecast error calculated based on the past actual values are used to calculate another power generation forecast value and another demand forecast value, the possibility of calculating an inappropriate forecast value as the other power generation forecast value and another demand forecast value can be reduced. Therefore, in order to ensure that the constraint conditions are satisfied for all of the multiple power generation forecast values and multiple demand forecast values during the optimization calculation, it is not necessary to unnecessarily relax the constraint conditions. As a result, it is possible to perform an optimization calculation that reduces the power procurement fee while satisfying the constraint conditions under appropriate constraint conditions. In other words, if the optimization calculation is performed using the multiple power generation forecast values and multiple demand forecast values acquired and calculated by the present invention, it is possible to create a charge/discharge plan that is applicable within the range that satisfies the appropriately set constraint conditions even if the actual power generation power and demand power deviate from the actual forecast values.

1 is a diagram illustrating an example of the overall configuration of a power management system according to an embodiment of the present invention. 2 is a block diagram showing an example of a functional configuration of a power management device according to the present embodiment. FIG. 5 is a diagram showing an example of power generation record data stored in a power generation record data storage unit; FIG. 11 is a diagram showing an example of demand result data stored in a demand result data storage unit. FIG. 10 is a diagram showing an example of vehicle usage record data stored in a vehicle usage record data storage unit; FIG. 4 is a block diagram showing a specific example of a functional configuration of a vehicle state prediction unit; FIG. 2 is a schematic diagram showing the concept of a prediction model used as an example by a vehicle usage state prediction unit; FIG. 11 is a schematic diagram for explaining the process of predicting a vehicle usage state for each predetermined unit time by beam search. FIG. 11 is a diagram for explaining a processing example of a consumption prediction value calculation unit. FIG. FIG. 13 is a diagram for explaining an example of calculation of a prediction error. 4 is a flowchart showing an example of the operation of the power management device according to the present embodiment. FIG. 13 is a diagram illustrating an example of gradual relaxation of constraint conditions. 11 is a block diagram showing another example of the functional configuration of the power management device according to the present embodiment. FIG.

An embodiment of the present invention will be described below with reference to the drawings. FIG. 1 is a diagram showing an example of the overall configuration of a power management system according to this embodiment. The power management system of this embodiment is configured to include a power management device 100 installed as a server device on the cloud, for example, a system (EMS: Energy Management System) 101 that manages the power used by consumers, and a V2H (Vehicle to Home) system 102 that enables the power stored in an on-board storage battery 113 to be used as home power.

A consumer is a person who receives and uses the supply of electric power, and in one example, it is a general household. A consumer may be a business establishment such as a factory, an office, a sales office, a school, or a hospital. In the facility of the consumer, an EMS 101, a V2H system 102, a photovoltaic power generation device 111 which is an example of a renewable energy power generation device, various power loads 112 _-1 , 112 _-2 , ... (hereinafter simply referred to as power loads 112) that consume electric power, an in-vehicle storage battery 113, and a non-in-vehicle storage battery 114 are set. In the case of a general household, the power load 112 is various electrical appliances such as lighting, a television, a refrigerator, a washing machine, and an air conditioner, and the storage battery 114 is a small residential storage battery. In the following, the residential storage battery 114 is described.

The EMS 101 is connected to a V2H system 102, a solar power generation device 111, a power load 112, and a residential storage battery 114. By using the EMS 101, it is possible to measure the power usage in the power load 112 in real time and visualize the energy usage status for the entire building, for each room, for each electrical appliance, and for each hour. It is also possible to perform "peak shifting" by shifting the peak to other time periods and leveling it out by using the power load 112 to avoid time periods with high power usage, and "peak cutting" by preventing power usage from exceeding a predetermined target value by optimally using the power generated by the solar power generation device 111 and the power stored in the vehicle storage battery 113 and the residential storage battery 114.

The vehicle's on-board storage battery 113 is detachably connected to the V2H system 102 by a dedicated cable. When the V2H system 102 and the on-board storage battery 113 are connected, it is possible to charge the on-board storage battery 113 with electricity generated by the solar power generation device 111 (hereinafter referred to as generated electricity) or electricity purchased from a power company through a grid not shown (hereinafter referred to as purchased electricity), or to discharge electricity stored in the on-board storage battery 113 (hereinafter referred to as stored electricity) for use by the power load 112.

On the other hand, when the vehicle is used as mobility (a means of transportation), the dedicated cable connecting the V2H system 102 and the on-board storage battery 113 is disconnected, allowing the vehicle to move freely. In this embodiment, the V2H system 102 determines that the vehicle is "at home" when the on-board storage battery 113 is connected by a dedicated cable, and that the vehicle is "out" (being used as mobility) when the on-board storage battery 113 is not connected by a dedicated cable.

The power management device 100 is connected to the EMS 101 via a communication network such as the Internet, and acquires various data via the EMS 101. The power management device 100 is also connected to the electricity rate provider server 200 via the communication network, and acquires data on the electricity rate system by time period from the electricity rate provider server 200. Then, using this acquired data, it executes processes such as various predictions and the creation of charge and discharge plans, which will be described later, and controls the EMS 101 based on the results.

2 is a block diagram showing an example of the functional configuration of the power management device 100 according to this embodiment. As shown in FIG. 2, the power management device 100 according to this embodiment includes, as its functional configuration, a power generation record data recording unit 1A, a power generation prediction unit 1B, a power generation prediction error calculation unit 1C, a separate power generation prediction value calculation unit 1D, a demand record data recording unit 2A, a demand power prediction unit 2B, a demand prediction error calculation unit 2C, a separate demand prediction value calculation unit 2D, a vehicle usage record data recording unit 3A, a vehicle state prediction unit 3B, a consumption prediction error calculation unit 3C, a separate consumption prediction value calculation unit 3D, a fee data acquisition unit 4, a charge/discharge plan creation unit 5, and a charge/discharge control unit 6. In addition, the power management device 100 according to this embodiment includes, as storage media, a power generation record data storage unit 11, a demand record data storage unit 12, a vehicle usage record data storage unit 13, and a fee table storage unit 14.

The above-mentioned functional blocks 1A-1D, 2A-2D, 3A-3D, 4-6 can be configured using either hardware, a DSP (Digital Signal Processor), or software. For example, when configured using software, the above-mentioned functional blocks are actually configured with a computer's CPU, RAM, ROM, etc., and are realized by the operation of a program stored in a storage medium such as the RAM, ROM, hard disk, or semiconductor memory.

The power management device 100, with the functional configuration shown in FIG. 2, creates a charge/discharge plan for the storage batteries 113, 114 based on the predicted power generation by the photovoltaic power generation device 111, the predicted power demand by the power load 112, the predicted power consumption by the on-board storage battery 113, the charge/discharge amount of the storage batteries 113, 114, and a time-of-day electricity rate system. In this embodiment, the charge/discharge amount of the storage batteries 113, 114 at each time during the prediction target period is one of the explanatory variables, and a mathematical optimization model function with the power procurement price through power purchase as the objective function is used to perform optimization calculations based on multiple predicted values to create a charge/discharge plan that minimizes the objective function while satisfying all specified constraint conditions.

The power generation performance data recording unit 1A acquires measurement data of the amount of power generated by the solar power generation device 111 from the EMS 101 in real time, and stores it in the power generation performance data storage unit 11. In this embodiment, the power generation performance data recording unit 1A acquires measurement data of the amount of power generated transmitted from the EMS 101 at predetermined unit times, and stores it in the power generation performance data storage unit 11. The power generation performance data recording unit 1A may also acquire the measurement data recorded by the EMS 101 at predetermined unit times in batches.

Figure 3 is a diagram showing an example of power generation performance data stored in the power generation performance data storage unit 11. As shown in Figure 3, the power generation performance data storage unit 11 stores the time, weather, and amount of power generation for each predetermined unit time (e.g., every 30 minutes) in chronological order. The weather information is obtained, for example, from a website or server that provides meteorological information via a communication network such as the Internet.

The demand record data recording unit 2A acquires measurement data of the demand power amount (power consumption amount) of the power load 112 from the EMS 101 in real time, and stores it in the demand record data storage unit 12. In this embodiment, the demand record data recording unit 2A acquires measurement data of the total power consumption amount of each power load 112 _-1 , 112 _-2 , ... transmitted from the EMS 101 at predetermined unit time intervals, and stores it in the demand record data storage unit 12. The demand record data recording unit 2A may acquire the measurement data at predetermined unit time intervals recorded by the EMS 101 in batches.

Figure 4 is a diagram showing an example of actual demand data stored in the actual demand data storage unit 12. As shown in Figure 4, the actual demand data storage unit 12 stores the time, day of the week, and amount of demanded power for each predetermined unit time (e.g., every 30 minutes) in chronological order. The day of the week is obtained, for example, from calendar data stored in advance in the power management device 100.

The vehicle usage history data recording unit 3A acquires detection data of the vehicle's usage state as a mobility (either at home or away from home) and measurement data of the power consumption of the vehicle-mounted storage battery 113 in real time from the EMS 101, and stores them in the vehicle usage history data storage unit 13. In this embodiment, the vehicle usage history data recording unit 3A acquires detection data of the vehicle usage state transmitted from the EMS 101 at predetermined unit times and stores them in the vehicle usage history data storage unit 13. The vehicle usage history data recording unit 3A also acquires measurement data of the power consumption transmitted when the EMS 101 detects that the vehicle usage state has changed from away to at home, and stores them in the vehicle usage history data storage unit 13. The power consumption of the vehicle-mounted storage battery 113 can be calculated from the difference between the stored power of the vehicle-mounted storage battery 113 measured when the vehicle usage state changes from at home to away from home (at the time of departure) and the stored power of the vehicle-mounted storage battery 113 measured when the vehicle usage state changes from away to at home (at the time of return). The vehicle usage history data recording unit 3A may also acquire the detection data of the usage state for each predetermined unit time recorded by the EMS 101 and the measurement data of the power consumption amount for each state change in a batch manner.

5 is a diagram showing an example of vehicle usage history data stored in the vehicle usage history data storage unit 13. As shown in FIG. 5, the vehicle usage history data storage unit 13 stores the time, day of the week, and vehicle usage status for each predetermined unit time (e.g., every 30 minutes) in chronological order, and stores the power consumption of the vehicle storage battery 113 when the vehicle is out. In FIG. 5, a circle in the vehicle usage status indicates that the vehicle is at home, and a cross indicates that the vehicle is out. In the example of FIG. 5, it is recorded that the vehicle was out for 7 hours and 30 minutes from 8:30 to 15:30 on 4/1/2019, and the power consumption of the vehicle storage battery 113 during that time. Since the vehicle was not out before 8:30 and after 16:00, the power consumption of the vehicle storage battery 113 is not recorded. For the power consumption for each predetermined unit time (every 30 minutes), for example, a value obtained by averaging the power consumption recorded as shown in FIG. 5 is used.

The power generation prediction unit 1B uses the actual power generation data stored in the actual power generation data storage unit 11 to predict the amount of power generated by the solar power generation device 111 for each specified unit time during the prediction period (corresponding to the "amount of power generated at each time during the prediction period" described in the claims. The amount of power generated at time t means the amount of power generated from time t to time t+1. This also applies to the amount of power demand and power consumption described below).

Here, the power generation prediction unit 1B repeats the prediction of the amount of power generated by the solar power generation device 111 at a predetermined cycle as a batch process. For example, the power generation prediction unit 1B predicts the amount of power generated at a set time each day (e.g., midnight) for a prediction period of 24 hours the following day. Note that the time and prediction period for this prediction are merely examples and are not limited to these.

The power generation prediction unit 1B uses the power generation performance data stored in the power generation performance data storage unit 11 to analyze the tendency of how much power generation can be obtained in what time period and under what weather conditions, and predicts the amount of power generation for each predetermined unit time during the prediction period based on the analysis results and the predicted weather information for each predetermined unit time during the prediction period. The predicted weather information is obtained, for example, from a website or server that provides meteorological information via a communication network such as the Internet.

As an example, the power generation prediction unit 1B uses the power generation performance data stored in the power generation performance data storage unit 11 to generate a prediction model consisting of a function with the time period and weather as explanatory variables and the amount of power generation as the objective variable. Then, by inputting predicted weather information for each predetermined unit time during the prediction period into the prediction model, the amount of power generation for each predetermined unit time during the prediction period is predicted. The prediction model can be generated, for example, using known machine learning. For example, the prediction model can be a regression prediction model using machine learning such as random forest.

The power demand prediction unit 2B predicts the amount of power demand at the power load 112 for each predetermined unit time during the prediction period (the amount of power demand at each time during the prediction period) using the actual demand data stored in the actual demand data storage unit 12. Here, the power demand prediction unit 2B repeatedly predicts the amount of power demand by the power load 112 as a batch process. For example, the power demand prediction unit 2B predicts the amount of power demand at a set time each day (for example, midnight) for the prediction period of 24 hours the following day. Note that the time at which this prediction is made and the prediction period are merely examples and are not limited to these.

The electricity demand forecasting unit 2B uses the actual demand data stored in the actual demand data storage unit 12 to analyze the tendency of how much electricity demand will occur during which time period on which day of the week, and forecasts the amount of electricity demand for each specified unit time during the forecast period based on the analysis results and information on the day of the week during the forecast period.

As an example, the power demand prediction unit 2B uses the actual demand data stored in the actual demand data storage unit 12 to generate a prediction model consisting of a function with the time period and day of the week as explanatory variables and the amount of demanded power as the objective variable. Then, by inputting information about the day of the week during the prediction period into the prediction model, the amount of demanded power for each predetermined unit time during the prediction period is predicted. The prediction model can be generated using, for example, known machine learning. For example, the prediction model can be a regression prediction model using machine learning such as random forest. Note that a prediction model may be generated that uses other information such as weather information (weather, temperature, humidity, etc.) as explanatory variables in addition to the time period and day of the week. In this case, the demand actual data stored in the actual demand data storage unit 12 also includes other information.

The vehicle state prediction unit 3B predicts the usage state of the vehicle as a mobility and the power consumption of the on-board storage battery 113 (the vehicle usage state and power consumption at each time of the prediction target period) for each predetermined unit time during the prediction target period using the vehicle usage history data of the on-board storage battery 113 stored in the vehicle usage history data storage unit 13. Here, the vehicle state prediction unit 3B repeatedly predicts the vehicle usage state and the power consumption of the on-board storage battery 113 as a batch process. For example, the vehicle state prediction unit 3B predicts the vehicle usage state and the power consumption at a set time each day (for example, midnight) for the prediction target period of 24 hours the following day. Note that the time and the prediction target period at which this prediction is made are merely examples and are not limited thereto.

The vehicle state prediction unit 3B uses the vehicle usage history data of the on-board storage battery 113 stored in the vehicle usage history data storage unit 13 to analyze the tendency of which days of the week and times of the day the vehicle is used as mobility and how much power is generated, and predicts the vehicle usage state and power consumption for each predetermined unit time during the prediction target period based on the analysis result and the day of the week information during the prediction target period. For example, the vehicle state prediction unit 3B predicts the vehicle usage state for each predetermined unit time during the prediction target period. Then, from the past vehicle usage states stored as vehicle usage history data in the vehicle usage history data storage unit 13, a vehicle usage state similar to the predicted vehicle usage state is extracted, and the power consumption of the on-board storage battery 113 is predicted based on the actual power consumption value corresponding to the extracted similar vehicle usage state.

Figure 6 is a block diagram showing a specific example of the functional configuration of the vehicle state prediction unit 3B. As shown in Figure 6, the vehicle state prediction unit 3B is configured with a vehicle usage state prediction unit 31, a pattern extraction unit 32, and a consumption prediction value calculation unit 33.

The vehicle usage state prediction unit 31 predicts multiple patterns of vehicle usage state as a form of mobility at each time during the prediction period, and calculates the occurrence probability that indicates the likelihood of each usage state occurring. The vehicle usage state prediction unit 31 predicts the vehicle usage state for each predetermined unit time during the prediction period, and calculates the occurrence probability of the predicted vehicle usage state, for example, by performing processing similar to that disclosed in JP 2022-2449 A.

That is, the vehicle usage status prediction unit 31 uses the vehicle usage history data stored in the vehicle usage history data storage unit 13 to generate a prediction model consisting of a function with the time period and day of the week as explanatory variables and the vehicle usage status as the objective variable. Then, by inputting information about the day of the week during the prediction period into the prediction model, the vehicle usage status for each specified unit time during the prediction period is predicted. Note that a prediction model may be generated that uses other information, such as weather information, as explanatory variables in addition to the time period and day of the week. In this case, the vehicle usage history data stored in the vehicle usage history data storage unit 13 also includes other information.

Fig. 7 is a schematic diagram showing the concept of a prediction model used as an example by the vehicle usage state prediction unit 31. The prediction model shown in Fig. 7 is a prediction model configured to input date and time information and day of the week information for each predetermined unit time to a neural network and output the vehicle usage state for each predetermined unit time from the neural network. The neural network includes a long short-term memory (LSTM) layer and a function layer between the input layer and the output layer, and predicts the vehicle usage state for a certain predetermined unit time _tn by inheriting the vehicle usage state (label) predicted for the previous predetermined unit time _tn-1 .

By using such a prediction model, it is possible to reduce the possibility that the vehicle usage prediction result by the vehicle usage prediction unit 31 will be an erroneous prediction in which the time period when the vehicle is predicted to be at home and the time period when the vehicle is predicted to be away alternate at short time intervals. In other words, alternating between being at home and being away at short time intervals of 30 minutes is not a typical usage state for a vehicle as a form of mobility, and both the time period when the vehicle is at home and the time period when the vehicle is away often last for a relatively long period of time. By assigning a label representing the predicted result of the vehicle usage state for each specified unit of time and generating a prediction model that applies sequence labeling using LSTM, it becomes possible to predict the vehicle usage state in accordance with general trends.

The vehicle usage prediction unit 31 inputs prediction period information (date and time information and day of the week information for the prediction period) into a prediction model conceptually described in FIG. 7, and predicts multiple patterns of vehicle usage as mobility for each specified unit of time during the prediction period while pruning some of the prediction results with low probability of occurrence from among multiple prediction results for each specified unit of time. When making this prediction, the vehicle usage prediction unit 31 uses, for example, a technique called beam search.

Figure 8 is a schematic diagram for explaining the process of predicting the vehicle usage state for each predetermined unit time by beam search. In Figure 8, time t1 is the first predetermined unit time of the prediction target period, time t2 is the second predetermined unit time of the prediction target period, and time t3 is the third predetermined unit time of the prediction target period. The fourth and subsequent predetermined unit times are not shown. In Figure 8, the ■ symbol indicates that the vehicle is predicted to be out, and the appended numerical value indicates the occurrence probability. The □ symbol indicates that the vehicle is predicted to be at home, and the appended numerical value indicates the occurrence probability.

In the example of FIG. 8, the usage state prediction unit 63 predicts that the probability of the vehicle usage state at time t1 is 0.8 for being out and 0.2 for being at home. Here, the vehicle usage state prediction unit 31 prunes, for example, prediction results other than those with the top N occurrence probabilities from among the two prediction results for the vehicle usage state at time t1. Pruning means not adopting them. Here, N=3, and a description will be given assuming that prediction results other than those with the top three occurrence probabilities are pruned. At time t1, only two patterns are predicted, so pruning is not performed.

Next, the vehicle usage state prediction unit 31 predicts that the probability of the vehicle usage state at time t2 remaining in the out state from the out state at time t1 is 0.5, the probability of changing from the out state at time t1 to the at-home state is 0.3, the probability of changing from the at-home state at time t1 to the out state is 0.05, and the probability of continuing in the at-home state from the at-home state at time t1 is 0.15. In this case, the vehicle usage state prediction unit 31 prunes the prediction result of the pattern "time t1: at-home → time t2: out", which is predicted to have the lowest occurrence probability of 0.05 among the four prediction results.

Next, the vehicle usage state prediction unit 31 predicts the vehicle usage state at time t3. Here, the vehicle usage state prediction unit 31 does not perform predictions after time t3 for the vehicle usage states pruned at time t2, but takes over the prediction results of the three surviving patterns at time t2 and predicts the probability of being out/at home for each of them. Therefore, six patterns of vehicle usage states are predicted as follows:
(1) Time t1: Going out → Time t2: Going out → Time t3: Going out... Probability 0.1
(2) Time t1: Out → Time t2: Out → Time t3: At home... Probability 0.4
(3) Time t1: out → Time t2: at home → Time t3: out ... probability 0.1
(4) Time t1: Out → Time t2: At home → Time t3: At home... Probability 0.2
(5) Time t1: At home → Time t2: At home → Time t2: Out ... Probability 0.12
(6) Time t1: At home → Time t2: At home → Time t2: At home... Probability 0.03
In this case, the vehicle usage state prediction unit 31 prunes patterns (1), (3), and (6) that are outside the top three occurrence probabilities.

The vehicle usage state prediction unit 31 then similarly predicts six patterns of vehicle usage state for each period from time t4 (not shown) to time t48, which is the last specified unit time of the prediction target period (the next 24 hours) (the final time when the specified unit time is 30 minutes), pruning the predicted results of patterns other than the top three in occurrence probability. Then, by pruning three patterns other than the top three in occurrence probability from the six predicted results at time t48, three patterns of predicted usage state are finally obtained.

The pruning method shown here is an example, and is not limited to this. For example, prediction results with an occurrence probability equal to or less than a predetermined value (for example, equal to or less than 0.1) may be pruned. In addition, although an example of predicting three patterns of vehicle use states has been described here, a greater number of patterns of vehicle use states may be predicted. Although not described in detail here, multiple patterns of vehicle use states may be predicted by the method described as Modification 1 or Modification 2 in JP2022-2449A. In addition, any method may be used as long as the occurrence probability is calculated for each pattern, and the method is not limited to the method described in JP2022-2449A.

The pattern extraction unit 32 extracts one usage state prediction pattern based on the occurrence probability from among the multiple usage state prediction patterns generated by the vehicle usage state prediction unit 31. For example, the pattern extraction unit 32 extracts the usage state prediction pattern with the highest occurrence probability.

The consumption prediction value calculation unit 33 calculates a prediction value of the power consumption of the on-board storage battery 113 at each time of the prediction period based on the usage state prediction pattern extracted by the pattern extraction unit 32 and the actual values of the past power consumption of the on-board storage battery 113 stored in the vehicle usage record data storage unit 13. Here, the consumption prediction value calculation unit 33 uses the actual values close to the departure time and the return home time to predict the power consumption of the on-board storage battery 113 for each outing.

Figure 9 is a diagram for explaining an example of the processing of the consumption prediction value calculation unit 33. Figure 9 (0) is a usage state prediction pattern extracted by the pattern extraction unit 32. First, as shown in Figure 9 (a), the consumption prediction value calculation unit 33 extracts actual usage state patterns whose departure time and return home time differ from the usage state prediction pattern in Figure 9 (0) by within one hour from actual usage state patterns for D days (D=5 in the example of Figure 9 (a)) prior to the prediction target period of 4/6.

Next, as shown in FIG. 9(b), the consumption prediction value calculation unit 33 extracts K actual usage state patterns (K=3 in the case of FIG. 9(b)) in descending order of the overlap with the predicted usage state pattern in terms of time. Finally, the consumption prediction value calculation unit 33 calculates the median of the power consumption when going out for the K actual usage state patterns, and sets this as the predicted value of the power consumption of the in-vehicle storage battery 113. If there is no actual usage state pattern that meets the first condition, the power consumption of the actual usage state pattern that has the greatest overlap with the predicted usage state pattern in terms of time among the actual usage state patterns for the past 30 days is set as the predicted value of the power consumption of the in-vehicle storage battery 113.

The power generation prediction error calculation unit 1C calculates the prediction error of the amount of power generation expected at each time of the prediction period (hereinafter referred to as the power generation prediction error) based on the actual values of the amount of power generation in the past of the photovoltaic power generation device 111 stored in the power generation performance data storage unit 11. That is, the power generation prediction error calculation unit 1C calculates the power generation prediction error for each of the times t1 to t48, which are 30 minutes apart during the prediction period. The power generation prediction error calculation unit 1C calculates at least one, and preferably multiple, power generation prediction errors for each time.

For example, the power generation prediction error calculation unit 1C calculates the standard deviation of the actual values of past power generation amounts, and generates a random value within a predetermined value range specified based on the standard deviation to determine the power generation prediction error. As an example, the power generation prediction error calculation unit 1C takes the power generation prediction value acquired by the power generation prediction unit 1B as the average, assumes a normal distribution from the standard deviation σ as shown in FIG. 10, and generates a random value within the range of ±2σ to determine the power generation prediction error. In this way, in this embodiment, errors with standard deviations exceeding the range of ±2σ, that is, errors that are unlikely to occur probabilistically, are not adopted.

The separate power generation forecast value calculation unit 1D calculates a separate power generation forecast value at each time of the prediction target period using the power generation forecast value acquired by the power generation power prediction unit 1B and the power generation forecast error calculated by the power generation forecast error calculation unit 1C. The separate power generation forecast value calculation unit 1D calculates at least one, and preferably multiple, separate power generation forecast values for each time.

For example, assuming that the power generation forecast value at each time t during the forecast period is PPV _t and the power generation forecast error expected at each time t is PPE _t , the separate power generation forecast value calculation unit 1D calculates the separate power generation forecast value PPV _t ' at each time t using the following (Equation 1): Here, rand(σ) is a function that generates a random number within the range of ±2σ according to a normal distribution with a mean of 0 and a standard deviation of σ.

The demand forecast error calculation unit 2C calculates the forecast error of the amount of power demand expected at each time of the prediction period (hereinafter, called the demand forecast error) based on the actual values of the past amount of power demand of the power load 112 stored in the demand actual data storage unit 12. That is, the demand forecast error calculation unit 2C calculates the demand forecast error for each of the times t1 to t48, which are 30 minutes apart during the prediction period. The demand forecast error calculation unit 2C calculates at least one, and preferably multiple, demand forecast errors for each time.

For example, the demand forecast error calculation unit 2C calculates the standard deviation of the actual values of the past demand energy amounts, and generates a random value within a predetermined value range specified based on the standard deviation to determine the demand forecast error. As an example, the demand forecast error calculation unit 2C, like the power generation forecast error calculation unit 1C, also assumes a normal distribution as shown in FIG. 10, and generates a random value within a range of ±2σ to determine the demand forecast error.

The separate demand forecast value calculation unit 2D calculates a separate demand forecast value at each time of the forecast period using the demand forecast value acquired by the power demand forecast unit 2B and the demand forecast error calculated by the demand forecast error calculation unit 2C. The separate demand forecast value calculation unit 2D calculates at least one, and preferably multiple, separate demand forecast values for each time. For example, assuming that the demand forecast value at each time t during the forecast period is DPV _t and the demand forecast error expected at each time t is DPE _t , the separate demand forecast value calculation unit 2D calculates a separate demand forecast value DPV _t ' for each time t using the following (Equation 2).

The consumption prediction error calculation unit 3C calculates the prediction error of the battery power consumption expected at each time of the prediction period (hereinafter referred to as the consumption prediction error) based on the actual values of the past power consumption of the on-board storage battery 113 stored in the vehicle usage data storage unit 13. That is, the consumption prediction error calculation unit 3C calculates the consumption prediction error for each of the times t1 to t48, which are 30 minutes apart during the prediction period. The consumption prediction error calculation unit 3C calculates at least one, and preferably multiple, consumption prediction errors for each time.

For example, the consumption prediction error calculation unit 3C calculates the standard deviation of the actual values of past power consumption of the storage battery, and generates a random value within a predetermined value range specified based on the standard deviation, which is used as the consumption prediction error. As an example, like the power generation prediction error calculation unit 1C, the consumption prediction error calculation unit 3C also assumes a normal distribution as shown in FIG. 10, and generates a random value within a range of ±2σ, which is used as the consumption prediction error.

The separate consumption prediction value calculation unit 3D calculates another battery consumption prediction value at each time of the prediction period using the battery consumption prediction value acquired by the vehicle state prediction unit 3B and the consumption prediction error calculated by the consumption prediction error calculation unit 3C. For example, assuming that the consumption prediction value at each time t during the prediction period is CPV _t and the consumption prediction error expected at each time t is CPE _t , the separate consumption prediction value calculation unit 3D calculates the separate consumption prediction value CPV _t ' at each time t using the following (Equation 3).

The fee data acquisition unit 4 acquires data on the electricity fee system by time period in the prediction period from the electricity fee provider server 200 via the communication network, and stores the data in the fee table storage unit 14. The electricity fee system data acquired here is, for example, data on the next day's trading price (contract price) provided by the Japan Electric Power Exchange (JEPX), and this can be treated as a predicted value in the prediction period. Note that a prediction service provided by a third party other than JEPX may be used to acquire predicted values of the trading price, or the fee data acquisition unit 4 may be provided with a function for predicting the trading price.

The charge/discharge plan creation unit 5 uses the predicted values calculated by the power generation prediction unit 1B, the separate power generation prediction value calculation unit 1D, the power demand prediction unit 2B, the separate demand prediction value calculation unit 2D, the vehicle state prediction unit 3B, and the separate consumption prediction value calculation unit 3D, as well as information on the electricity rate system by time period during the prediction target period stored in the rate table storage unit 14, to perform optimization calculations that reduce the electricity procurement fee for purchasing electricity while satisfying specified constraints and without causing a shortage of the amount of stored electricity required for use as vehicle mobility, thereby creating a charge/discharge plan for the storage batteries 113, 114.

Here, the respective predicted values used by the charge/discharge plan creation unit 5 are a plurality of power generation predicted values PPV _t , PPV _t ' obtained by the power generation prediction unit 1B and the separate power generation predicted value calculation unit 1D, a plurality of demand predicted values DPV _t , DPV _t ' obtained by the power demand prediction unit 2B and the separate demand predicted value calculation unit 2D, a plurality of consumption predicted values CPV _t , CPV _t ' obtained by the consumption predicted value calculation unit 33 and the separate consumption predicted value calculation unit 3D of the vehicle state prediction unit 3B, and the vehicle usage state predicted by the vehicle usage state prediction unit 31 of the vehicle state prediction unit 3B.

The charge/discharge plan creation unit 5 creates a charge/discharge plan for each predetermined unit time during the prediction target period for the storage batteries 113, 114 under predetermined constraint conditions so that the above-mentioned predicted values and the states of the storage batteries 113, 114 satisfy the constraint conditions, while reducing the electricity procurement fee for purchasing electricity without causing a shortage of the amount of stored electricity required for use as the vehicle's mobility.

For example, if the power procurement fee for purchasing power is represented as PPF, the charge amount of the storage batteries 113, 114 at each time t in the prediction period as CAT _t , the discharge amount of the storage batteries 113, 114 at each time t as DAT _t , multiple power generation forecast values at each time t as PPV _t,p , multiple demand forecast values at each time t as DPV _tp , and the electricity price at each time t as EP _t , the charge/discharge plan creation unit 5 uses the following (Equation 4) as an objective function and performs optimization calculations to create a charge/discharge plan that minimizes the power procurement fee PPF while satisfying all predetermined constraints. In (Equation 4), the charge amount CAT _t and discharge amount DAT _t of the storage batteries 113, 114 at each time t are operation variables (optimization variables) optimized by the optimization calculation.

Here, the symbol "p" used in the multiple power generation forecast values PPV _t,p and multiple demand forecast values DPV _tp indicates a pattern number of the forecast value. For example, the power generation forecast value PPV _t,1 of p=1 is the actual power generation forecast value PPV _t predicted by the power generation forecast unit 1B. Moreover, the power generation forecast value PPV _t,p of p=2, 3, ... is another power generation forecast value PPV _t ' calculated by the separate power generation forecast value calculation unit 1D. Similarly, the demand forecast value DPV _t,1 of p=1 is the actual demand forecast value DPV _t predicted by the power demand forecast unit 2B. Moreover, the demand forecast value DPV _t,p of p=2, 3, ... is another demand forecast value DPV _t ' calculated by the separate demand forecast value calculation unit 2D.

Note that minimizing the power procurement fee PPF through optimization calculations is not limited to strictly determining the actual minimum value. For example, a limit may be set on the number of times the recursive calculations are performed in consideration of the calculation load, and the power procurement fee PPF may be minimized within that limit. Alternatively, the recursive calculations may be performed until the power procurement fee PPF falls below a desired threshold, and the solution may be the one that has been minimized at the point when the power procurement fee PPF falls below the desired threshold.

The following describes the predetermined constraint conditions used by the charge/discharge plan creation unit 5 in the optimization calculation. In this embodiment, the predetermined constraint conditions include a plurality of constraint conditions set regarding the power generation forecast value, the demand forecast value, and the state of the storage batteries 113, 114 (charge/discharge amount and retained power). The plurality of constraint conditions include peak value-related constraint conditions for keeping the peak value of the power usage due to purchased power below a target peak value, and battery degradation-related constraint conditions for suppressing degradation of the storage batteries 113, 114 due to charging and discharging.

The following (Equation 5) is an example of a peak-related constraint: where PT is the peak target value.
( _DPVt,p + _CATt )-(PPVt _,p + _DATt )≦PT (Equation 5)
In this (Equation 5), the first term in the first parentheses on the left side indicates the amount of power usage at each time t, and the second term indicates the amount of power procurement other than power purchase at each time t. Therefore, the left side indicates the amount of power procured and used through power purchase. Generally, electricity charges consist of a basic charge and a metered charge, and the basic charge is determined by the maximum annual peak value of power usage. Therefore, a constraint is set to keep the amount of power usage through power purchase below the peak target value PT.

The following (Equation 6) to (Equation 11) are examples of battery degradation-related constraint conditions. Here, VCAT _t in (Equation 6) is the charge amount of the in-vehicle storage battery 113 from time t to time t+1, VDAT _t in (Equation 7) is the discharge amount of the in-vehicle storage battery 113 from time t to time t+1, RCAT _t in (Equation 8) is the charge amount of the residential storage battery 114 from time t to time t+1, and RDAT _t in (Equation 9) is the discharge amount of the residential storage battery 114 from time t to time t+1.
Lower limit value≦ _VCATt ≦upper limit value (Equation 6)
Lower limit value≦VDAT _t ≦upper limit value (Equation 7)
Lower limit value≦RCAT _t ≦upper limit value (Equation 8)
Lower limit value ≦ RDAT _t ≦ upper limit value (Equation 9)
Minimum value≦VBHP _t,p ≦Maximum value (Equation 10)
Minimum value≦RBHP _t ≦Maximum value (Equation 11)

Moreover, VBHP _t,p in the above (Equation 10) is the retained power of the in-vehicle storage battery 113 at each time t, and is calculated by the following (Equation 12). In this (Equation 10), the symbol "p" indicates a pattern number. For example, the retained power VBHP _t,1 for p=1 is the retained power calculated using the actual predicted consumption value CPV _t predicted by the vehicle state prediction unit 3B. Furthermore, the retained power VBHP _t,p for p=2, 3, ... is the retained power calculated using another predicted consumption value CPV _t ' calculated by the another predicted consumption value calculation unit 3D. RBHP _t in the above (Equation 11) is the retained power of the residential storage battery 114 at each time t, and is calculated by the following (Equation 13).

In the above (Equation 12), VBP ₀ is the initial held power of the in-vehicle storage battery 113 at the first time t0 of the prediction target period. FLG _t is a flag value indicating the vehicle use state predicted by the vehicle use state prediction unit 31, and is "1" when the user is out and "0" when the user is at home. In addition, in the above (Equation 13), RBP ₀ is the initial held power of the residential storage battery 114 at the first time t0 of the prediction target period.

The charge/discharge plan creation unit 5 performs optimization calculations by applying the peak value-related constraint condition shown in (Equation 5) above and the battery degradation-related constraint conditions shown in (Equations 6) to (Equations 11) as well as the constraint conditions shown below, for example.
Use of the power generated by the photovoltaic power generation device 111 for the power load 112 is prioritized over power purchased from the grid.
When the power generated by the solar power generation device 111 exceeds the power consumed by the power load 112 (i.e., when surplus power is generated), charging of the storage batteries 113, 114 with the surplus power is prioritized over charging with purchased power.
During a time period when it is predicted that the vehicle will be at home, charging and discharging is performed with priority given to the in-vehicle storage battery 113 over the residential storage battery 114.
Ensure that the stored power of the onboard battery 113 at the time when the vehicle usage state is predicted to switch from at home to away is not below the predicted value of the amount of power consumed while the vehicle is away until the next time when the vehicle usage state is predicted to switch from away to at home.

The following constraints may also be applied:
Considering emergency use of the vehicle for mobility, if the remaining charge of the on-board storage battery 113 at the time of returning home (when the on-board storage battery 113 is connected to the V2H system 102 via a dedicated cable) is below a threshold, charging will be performed up to a predetermined amount regardless of the vehicle usage state predicted by the vehicle usage state prediction unit 31.
- If the vehicle-mounted battery 113 is charged and discharged immediately after returning home after using the vehicle as a form of mobility and while the battery is still hot, battery deterioration will be rapid. Therefore, charging and discharging are performed after a predetermined time has elapsed since the vehicle has been used as a form of mobility and while the vehicle-mounted battery 113 is still hot.

The charge/discharge control unit 6 controls the actual charging/discharging of the storage batteries 113, 114 during the planned period (= prediction period) in accordance with the charge/discharge plan created by the charge/discharge plan creation unit 5 under the above-mentioned constraints.

Figure 11 is a flowchart showing an example of the operation of the power management device 100 according to this embodiment configured as described above. Note that the processes of the power generation record data recording unit 1A, the demand record data recording unit 2A, and the vehicle usage record data recording unit 3A are executed sequentially, and the processes of storing the respective record data in the power generation record data storage unit 11, the demand record data storage unit 12, and the vehicle usage record data storage unit 13 are executed separately and asynchronously from the flowchart shown in Figure 11.

In FIG. 11, the power generation prediction unit 1B predicts the amount of power generation at each time during the prediction period using the power generation performance data stored in the power generation performance data storage unit 11 (step S1). The power generation prediction error calculation unit 1C calculates the power generation prediction error expected at each time during the prediction period based on the past power generation performance values stored in the power generation performance data storage unit 11 (step S2). The separate power generation prediction value calculation unit 1D calculates another power generation prediction value at each time during the prediction period using the power generation prediction value acquired by the power generation prediction unit 1B and the power generation prediction error calculated by the power generation prediction error calculation unit 1C (step S3).

Next, the power demand forecasting unit 2B forecasts the amount of power demand at each time during the forecast period using the actual demand data stored in the actual demand data storage unit 12 (step S4). The demand forecast error calculation unit 2C calculates the demand forecast error expected at each time during the forecast period based on the actual values of the amount of power demand in the past stored in the actual demand data storage unit 12 (step S5). The separate demand forecast value calculation unit 2D then calculates another demand forecast value at each time during the forecast period using the demand forecast value acquired by the power demand forecasting unit 2B and the demand forecast error calculated by the demand forecast error calculation unit 2C (step S6).

Furthermore, the vehicle state prediction unit 3B predicts the usage state of the vehicle as a mobility and the power consumption of the on-board storage battery 113 at each time during the prediction target period using the vehicle usage history data of the on-board storage battery 113 stored in the vehicle usage history data storage unit 13 (step S7). Furthermore, the consumption prediction error calculation unit 3C calculates the consumption prediction error expected at each time during the prediction target period based on the past power consumption actual values of the on-board storage battery 113 stored in the vehicle usage history data storage unit 13 (step S8). Then, the separate consumption prediction value calculation unit 3D calculates another consumption prediction value at each time during the prediction target period using the consumption prediction value acquired by the vehicle state prediction unit 3B and the consumption prediction error calculated by the consumption prediction error calculation unit 3C (step S9).

The fee data acquisition unit 4 also acquires data on the electricity fee system by time period in the prediction target period from the electricity fee provider server 200, and stores the data in the fee table storage unit 14 (step S10). Note that the processes of steps S1 to S3, steps S4 to S6, steps S7 to S9, and step S10 do not necessarily have to be performed in this order.

Then, the charge/discharge plan creation unit 5 uses the predicted values calculated by the power generation prediction unit 1B, the separate power generation prediction value calculation unit 1D, the power demand prediction unit 2B, the separate demand prediction value calculation unit 2D, the vehicle state prediction unit 3B, and the separate consumption prediction value calculation unit 3D in the above steps S1 to S10, and the information on the electricity rate system by time period during the prediction target period stored in the rate table storage unit 14 to perform optimization calculations that reduce the electricity procurement fee for purchasing electricity while satisfying specified constraints and without causing a shortage of the amount of storage required for use as the mobility of the vehicle (step S11). This ends the processing of the flowchart shown in FIG. 11.

As described above in detail, in this embodiment, the power generation forecast value (the forecast value of the amount of power generation) of the photovoltaic power generation device 111 at each time of the prediction target period is obtained, and another power generation forecast value is calculated. In addition, the demand forecast value (the forecast value of the amount of demanded power) of the power load 112 at each time of the prediction target period is obtained, and another demand forecast value is calculated. In addition, the consumption forecast value (the forecast value of the amount of power consumption) of the storage batteries 113, 114 at each time of the prediction target period is obtained, and another consumption forecast value is calculated. Then, using information on the electricity rate system by time period in the prediction target period, under the constraint conditions set for the power generation forecast value, the demand forecast value, and the state of the storage battery, the multiple power generation forecast values, multiple demand forecast values, and multiple consumption forecast values acquired and calculated as described above satisfy all the constraint conditions, and the state of the storage battery reduces the power procurement fee for purchasing electricity without causing a shortage of the amount of storage required for use as the mobility of the vehicle. By performing optimization calculations, a charge and discharge plan for the storage batteries 113, 114 is created.

According to the present embodiment configured as described above, optimization calculations are performed using multiple power generation forecast values, multiple demand forecast values, and multiple consumption forecast values, including not only the actual power generation forecast value, demand forecast value, and consumption forecast value acquired initially, but also other power generation forecast values, demand forecast values, and consumption forecast values calculated based on the acquired actual power generation forecast value, demand forecast value, and consumption forecast value. In other words, optimization calculations are performed such that the multiple power generation forecast values, multiple demand forecast values, multiple consumption forecast values, and the state of the storage battery all satisfy the constraint conditions, and the power procurement fee for purchasing power is reduced without causing a shortage of the amount of storage required for use as the mobility of the vehicle. As a result, even if the actual power generation forecast, demand forecast, and consumption forecast are incorrect, it is highly likely that the constraint conditions will be satisfied, and it is possible to create an optimal charging and discharging plan that reduces the power procurement fee for purchasing power as much as possible without causing a shortage of the amount of storage required for use as the mobility of the vehicle.

In particular, in this embodiment, a different predicted value is calculated using the actual predicted value and a prediction error calculated based on past actual values, so that the possibility of calculating an inappropriate predicted value as the different predicted value can be reduced. Therefore, in order to ensure that the constraint conditions are satisfied for all of the multiple predicted values during the optimization calculation, it is not necessary to unnecessarily relax the constraint conditions, such as by increasing the peak target value PT in (Equation 5), decreasing the lower limit or minimum value of (Equation 6) to (Equation 11), or increasing the upper limit or maximum value of (Equation 6) to (Equation 11). This makes it possible to perform optimization calculations that reduce electricity procurement fees while satisfying appropriate constraint conditions.

Moreover, in this embodiment, the standard deviation σ is calculated based on past actual values, and another predicted value is calculated using only the prediction error extracted from the range of ±2σ of the normal distribution, so that it is possible to avoid the calculation of another predicted value, which is unlikely to occur probabilistically. In this way, by performing optimization calculations using multiple predicted values acquired and calculated by this embodiment, it is possible to create a charge/discharge plan that is applicable within the range that satisfies appropriately set constraint conditions, even if the actual generated power, demanded power, and consumed power deviate from the actual predicted values.

When optimization calculations are performed under appropriately set constraint conditions (constraint conditions that are not unnecessarily loose) as described above, a solution that satisfies all of the constraint conditions may not be obtained. In this case, the charge/discharge plan creation unit 5 may relax at least one of the constraint conditions and perform optimization calculations again under the relaxed constraint conditions to reduce the power procurement fee.

In this case, for example, the charge/discharge plan creation unit 5 may assign a higher priority to the peak value-related constraint conditions than the battery degradation-related constraint conditions, and repeat the optimization calculation while gradually relaxing the multiple constraint conditions according to the priority until a solution that satisfies the constraint conditions is obtained. For example, if a solution that satisfies all of the multiple constraint conditions shown in (Equation 5) to (Equation 11) above cannot be obtained, the battery degradation-related constraint conditions are first relaxed and the optimization calculation is performed again. Then, if a solution that satisfies all of the multiple constraint conditions including the relaxed battery degradation-related constraint conditions is still not obtained, the peak value-related constraint conditions are relaxed and the optimization calculation is performed again. When relaxing the peak value-related constraint conditions, it is preferable to restore the battery degradation-related constraint conditions.

Here, relaxing the battery degradation-related constraint conditions means relaxing one or more of the constraint conditions shown in (Equation 6) to (Equation 11) above. That is, relaxing one or more of the lower limit values or minimum values shown in (Equation 6) to (Equation 11) means reducing one or more of the lower limit values or minimum values shown in (Equation 6) to (Equation 11), increasing one or more of the upper limit values or maximum values shown in (Equation 6) to (Equation 11), or both. In addition, relaxing the peak value-related constraint conditions means increasing the peak target value PT shown in (Equation 5).

Here, an example of two-stage constraint relaxation in which battery degradation-related constraints are relaxed first and then peak value-related constraints are relaxed second has been described, but constraints may be relaxed in stages over three or more stages. Figure 12 is a diagram showing an example of relaxation in this case. In the example of Figure 12, first, only the battery degradation-related constraints are relaxed by one stage from the previous stage, the battery degradation-related constraints are restored to their previous state and the peak value-related constraints are relaxed by one stage from the previous stage, and the battery degradation-related constraints and the peak value-related constraints are relaxed by one stage each, which are considered as one set, and these sets are repeated in sequence to relax the constraints in stages. Note that this is merely one example, and other methods may be used.

In addition to the constraints shown in (Equation 6) to (Equation 11), a condition that a certain time period must elapse after returning home before charging or discharging the battery may be added as a battery degradation-related constraint to be relaxed. In this case, the relaxation of the condition is either to shorten the certain time period or to eliminate this constraint.

In the above embodiment, the configuration is shown to include both the vehicle storage battery 113 and the residential storage battery 114 as storage batteries, but it is also possible to include only one of them. In the case of a configuration including only the residential storage battery 114, the vehicle usage history data recording unit 3A, the vehicle state prediction unit 3B, the consumption prediction error calculation unit 3C, the separate consumption prediction value calculation unit 3D, and the vehicle usage history data storage unit 13 are omitted. In addition, the constraint conditions shown in (Equation 6), (Equation 7), and (Equation 10) are excluded. Then, the charge/discharge plan creation unit 5 creates a charge/discharge plan for the residential storage battery 114 by performing optimization calculations to reduce the power procurement fee under the multiple constraint conditions shown in (Equation 5), (Equation 8), (Equation 9), and (Equation 11).

In the above embodiment, an example has been described in which one usage state prediction pattern with the highest occurrence probability is extracted by the pattern extraction unit 32 of the vehicle state prediction unit 3B, one consumption prediction value is calculated by the consumption prediction value calculation unit 33 using the usage state prediction pattern, and another consumption prediction value is calculated by the other consumption prediction value calculation unit 3D using the consumption prediction error calculated by the consumption prediction error calculation unit 3C, but the present invention is not limited to this. For example, it may be configured as follows.

That is, the pattern extraction unit 32 extracts a plurality of usage state prediction patterns based on the occurrence probability from among the plurality of usage state prediction patterns generated by the vehicle usage state prediction unit 31. For example, the plurality of usage state prediction patterns are extracted in descending order of occurrence probability. The consumption prediction value calculation unit 33 calculates a prediction value of the power consumption of the on-board storage battery 113 at each time of the prediction target period for each of the plurality of usage state prediction patterns extracted by the pattern extraction unit 32, based on the usage state prediction pattern and the actual value of the past power consumption of the on-board storage battery 113.

In this configuration, the consumption prediction error calculation unit 3C calculates one or more consumption prediction errors for each of the multiple consumption prediction values, and the separate consumption prediction value calculation unit 3D calculates another consumption prediction value using the consumption prediction errors. As an example, m (m≧2) usage state prediction patterns with the highest occurrence probability may be extracted, and _n1 , _n2 , ..., _nm consumption prediction values (a total of _n1 + _n2 + ... + _nm ) may be calculated from each of the extracted m usage state prediction patterns. Here, _n1 > _n2 > ... > _nm may be set according to the occurrence probability of the usage state prediction pattern.

Alternatively, the consumption prediction error calculation unit 3C and the separate consumption prediction value calculation unit 3D may be omitted, and the charge/discharge plan creation unit 5 may be configured as follows. That is, under constraint conditions set for the power generation prediction value, the demand prediction value, and the state of the storage batteries 113, 114, the charge/discharge plan creation unit 5 creates a charge/discharge plan for the storage batteries 113, 114 by performing an optimization calculation that reduces the power procurement fee without causing a shortage of the amount of storage required for use as the mobility of the vehicle, while satisfying the constraint conditions among the multiple power generation prediction values obtained by the power generation prediction unit 1B and the separate power generation prediction value calculation unit 1D, the multiple demand prediction values obtained by the power demand prediction unit 2B and the separate demand prediction value calculation unit 2D, the multiple consumption prediction values obtained by the consumption prediction value calculation unit 33 from each of the multiple usage state prediction patterns extracted by the pattern extraction unit 32, and the state of the storage batteries 113, 114.

In addition, in the above embodiment, an example was described in which the fee data acquisition unit 4 acquires data on the electricity fee system by time period in the prediction target period from the electricity fee provider server 200, stores it in the fee table storage unit 14, and performs optimization calculations using this electricity fee system data, but the present invention is not limited to this. For example, similar to the power generation forecast value and demand forecast value, one or more separate electricity fees may be calculated using the prediction error of the electricity fee, and then the optimization calculations may be performed using the multiple electricity fees thus obtained.

Figure 13 is a diagram showing an example of the functional configuration of the power management device 100' in this case. In this Figure 13, blocks with the same reference numbers as those in Figure 2 include the same functions as the functional blocks shown in Figure 2, and are shown in a simplified manner. The power management device 100' shown in Figure 13 includes an electricity rate actual data recording unit 4A, a rate data acquisition unit 4B, an electricity rate prediction error calculation unit 4C, a separate electricity rate prediction value calculation unit 4D, an electricity rate actual data storage unit 14', and a charge/discharge plan creation unit 5' instead of the rate data acquisition unit 4, rate table storage unit 14, and charge/discharge plan creation unit 5 shown in Figure 2.

The electricity rate actual data recording unit 4A acquires actual data on the electricity rate system by time period from the electricity rate provider server 200 via the communication network, and stores the data in the electricity rate actual data storage unit 14'. The actual data on the electricity rate system acquired here is, for example, data on transaction prices (contract prices) previously determined by JEPX.

The fee data acquisition unit 4B corresponds to the electricity fee prediction unit in the claims, and similar to the fee data acquisition unit 4 shown in FIG. 2, acquires electricity fee data at each time of the prediction period from the electricity fee provider server 200 via a communication network. As described above, the electricity fee data acquired here is the next day's transaction price data provided by JEPX, and this can be treated as a predicted value for the prediction period (hereinafter referred to as the electricity fee prediction value).

The electricity rate prediction error calculation unit 4C calculates the prediction error of the electricity rate expected at each time of the prediction period (hereinafter referred to as the electricity rate prediction error) based on the past actual electricity rate values stored in the electricity rate actual data storage unit 14'. That is, the electricity rate prediction error calculation unit 4C calculates the electricity rate prediction error for each of the times t1 to t48, which are 30 minutes apart during the prediction period. The electricity rate prediction error calculation unit 4C calculates at least one, and preferably multiple, electricity rate prediction errors for each time.

The separate electricity price prediction value calculation unit 4D calculates a separate electricity price prediction value at each time of the prediction target period using the electricity price prediction value acquired by the price data acquisition unit 4B and the electricity price prediction error calculated by the electricity price prediction error calculation unit 4C. The separate electricity price prediction value calculation unit 4D calculates at least one, and preferably multiple, separate electricity price prediction values for each time.

The charge/discharge plan creation unit 5' uses a plurality of electricity price prediction values obtained by the price data acquisition unit 4B and the separate electricity price prediction value calculation unit 4D to perform optimization calculations to reduce the electricity procurement fee for purchasing electricity without causing a shortage of the amount of electricity storage required for use as vehicle mobility, while satisfying the constraint conditions set for the power generation prediction value, demand prediction value, and state of the storage batteries 113, 114, and the prediction values calculated by the power generation prediction unit 1B, separate power generation prediction value calculation unit 1D, power demand prediction unit 2B, separate demand prediction value calculation unit 2D, vehicle state prediction unit 3B, and separate consumption prediction value calculation unit 3D and the state of the storage batteries 113, 114.

In the above embodiment, regarding the battery degradation-related constraint condition of (Equation 10) and (Equation 11) that the stored power of the storage batteries 113, 114 is kept above a predetermined lower limit value and below a predetermined upper limit value, the lower limit value set for a predetermined time period in the prediction target period may be set to be larger than the lower limit values set for other time periods. The predetermined time period is set to be a time period in the latter half of the prediction target period. In this way, it is possible to prevent the creation of an unnatural charging and discharging plan that reduces the objective function by discharging in the latter half of the prediction target period. It is also possible to have the storage battery 113 serve as a safety margin by increasing the stored power before the vehicle starts to be used as a mobility vehicle.

In addition, in the above embodiment, when the charge/discharge plan creation unit 5 performs optimization calculations, with regard to the battery degradation-related constraint conditions of (Equation 10) and (Equation 11) that the retained power of the storage batteries 113, 114 is kept above a predetermined lower limit value and below a predetermined upper limit value, even if the retained power of the storage batteries 113, 114 falls below the lower limit value, if the period during which it falls below the lower limit value is within a predetermined time, the current retained power may be treated as the lower limit value during the predetermined time period.

In the above embodiment, an example of executing batch processing at the start of the prediction period has been described, but the present invention is not limited to this. For example, batch processing may be executed a predetermined time before the start of the prediction period. In addition, the user may be allowed to specify the timing of executing the batch processing.

In the above embodiment, a Home Energy Management System (HEMS) is used as an example of EMS 101, and an example of predicting whether a vehicle is at home or away is described, but the present invention is not limited to this. For example, the present invention can be similarly applied to xEMSs such as a Building Energy Management System (BEMS) and a Factory Energy Management System (FEMS).

In addition, the above embodiments are merely examples of how the present invention can be implemented, and the technical scope of the present invention should not be interpreted in a limiting manner. In other words, the present invention can be implemented in various forms without departing from its gist or main features.

Description of the Reference Number 1A Power generation result data recording unit 1B Power generation prediction unit 1C Power generation prediction error calculation unit 1D Individual power generation prediction value calculation unit 2A Demand result data recording unit 2B Demand power prediction unit 2C Demand prediction error calculation unit 2D Individual demand prediction value calculation unit 3A Vehicle usage result data recording unit 3B Vehicle state prediction unit 3C Consumption prediction error calculation unit 3D Individual consumption prediction value calculation unit 4 Fee data acquisition unit 4A Electricity fee result data recording unit 4B Fee data acquisition unit 4C Electricity fee prediction error calculation unit 4D Individual electricity fee prediction value calculation unit 5, 5' Charging and discharging plan creation unit 6 Charging and discharging control unit 31 Vehicle usage state prediction unit 32 Pattern extraction unit 33 Consumption prediction value calculation unit 100, 100' Power management device 101 EMS
102 V2H system 111 Photovoltaic power generation device 112 Power load 113 Vehicle storage battery 114 Residential storage battery

Claims (13)

A power management system that creates a charge/discharge plan for a storage battery based on a predicted power generation by a power generation device, a predicted power demand by a power load, a charge/discharge amount of the storage battery, and a time-of-day electricity rate system,
a power generation prediction unit for acquiring a predicted value of a power generation amount at each time of a prediction target period;
a power generation prediction error calculation unit that calculates a prediction error of the amount of power generation expected at each time of the prediction target period based on a past actual value of the amount of power generation;
a separate power generation forecast value calculation unit that calculates a separate power generation forecast value at each time of the prediction target period by using the power generation forecast value acquired by the power generation power prediction unit and the power generation forecast error calculated by the power generation forecast error calculation unit;
an electricity demand prediction unit that obtains a predicted value of an electricity demand amount at each time of the prediction target period;
a demand forecast error calculation unit that calculates a forecast error of the amount of energy demand expected at each time of the forecast target period based on an actual value of the amount of energy demand in the past;
a separate demand forecast value calculation unit that calculates a separate demand forecast value at each time of the prediction period by using the demand forecast value acquired by the power demand forecasting unit and the demand forecast error calculated by the demand forecast error calculation unit; and
and a charge/discharge plan creation unit that creates a charge/discharge plan for the storage battery by performing an optimization calculation using information on the electricity rate system for each time period during the prediction period under constraint conditions set for the power generation forecast value, the demand forecast value, and the state of the storage battery to reduce the power procurement fee for purchasing electricity while satisfying the constraint conditions between the multiple power generation forecast values obtained by the power generation forecast unit and the separate power generation forecast value calculation unit, the multiple demand forecast values obtained by the power demand forecast unit and the separate demand forecast value calculation unit, and the state of the storage battery. the power generation prediction error calculation unit calculates a standard deviation of the actual value of the past power generation amount, and generates a random value within a predetermined value range specified based on the standard deviation, as the power generation prediction error;
The power management system described in claim 1, characterized in that the demand forecast error calculation unit calculates a standard deviation of the actual values of the past demand energy, and generates a random value within a predetermined value range specified based on the standard deviation as the demand forecast error. The storage battery includes an on-board storage battery mounted in a vehicle,
a vehicle state prediction unit that obtains a predicted value of a usage state of the vehicle as a mobility and a predicted value of a power consumption amount of the vehicle-mounted storage battery at each time of the prediction target period;
a consumption prediction error calculation unit that calculates a prediction error of the power consumption expected at each time point during the prediction target period based on a past actual value of the power consumption of the vehicle-mounted storage battery;
a separate consumption prediction value calculation unit that calculates a separate consumption prediction value at each time of the prediction target period by using the consumption prediction value acquired by the vehicle state prediction unit and the consumption prediction error calculated by the consumption prediction error calculation unit,
2. The power management system according to claim 1, wherein the charge/discharge plan creation unit creates a charge/discharge plan for the storage battery by performing an optimization calculation to reduce the power procurement fee without causing a shortage of the amount of storage required for use of the vehicle as mobility, while the multiple power generation forecast values obtained by the power generation forecast unit and the separate power generation forecast value calculation unit, the multiple demand forecast values obtained by the power demand forecast unit and the separate demand forecast value calculation unit, the multiple consumption forecast values obtained by the vehicle state prediction unit and the separate consumption forecast value calculation unit, and the state of the storage battery satisfy the constraint conditions under constraint conditions set for the power generation forecast value, the demand forecast value, and the state of the storage battery. The vehicle state prediction unit is
a vehicle usage state prediction unit that predicts a plurality of patterns of usage states of the vehicle as a form of mobility at each time of the prediction target period and calculates an occurrence probability indicating a likelihood of each usage state pattern occurring;
a pattern extraction unit that extracts one usage state prediction pattern from the plurality of usage state prediction patterns generated by the vehicle usage state prediction unit based on the occurrence probability;
and a consumption prediction value calculation unit that calculates a predicted value of the power consumption of the on-board storage battery at each time of the prediction period based on the usage state prediction pattern extracted by the pattern extraction unit and actual values of past power consumption of the on-board storage battery. The vehicle state prediction unit is
a vehicle usage state prediction unit that predicts a plurality of patterns of usage states of the vehicle as a form of mobility at each time of the prediction target period and calculates an occurrence probability indicating the likelihood of each pattern occurring;
a pattern extraction unit that extracts a plurality of usage state prediction patterns from the plurality of usage state prediction patterns generated by the vehicle usage state prediction unit based on the occurrence probability;
4. The power management system according to claim 3, further comprising a consumption prediction value calculation unit that calculates, for each of a plurality of usage state prediction patterns extracted by the pattern extraction unit, a predicted value of the power consumption of the on-board storage battery at each time of the prediction period based on the usage state prediction pattern and actual values of past power consumption of the on-board storage battery. an electricity rate prediction unit that obtains a predicted value of the electricity rate at each time of the prediction target period;
an electricity price prediction error calculation unit that calculates a prediction error of the electricity price expected at each time of the prediction target period based on actual values of past electricity prices;
Further, a separate electricity price prediction value calculation unit calculates a separate electricity price prediction value at each time of the prediction target period using the electricity price prediction value acquired by the electricity price prediction unit and the electricity price prediction error calculated by the electricity price prediction error calculation unit,
2. The power management system according to claim 1, wherein the charge/discharge plan creation unit creates a charge/discharge plan for the storage battery by performing an optimization calculation using the plurality of electricity rate forecast values obtained by the electricity rate prediction unit and the separate electricity rate forecast value calculation unit, under constraint conditions set regarding the power generation forecast value, the demand forecast value, and the state of the storage battery, in which the plurality of power generation forecast values obtained by the power generation prediction unit and the separate power generation forecast value calculation unit, the plurality of demand forecast values obtained by the power demand prediction unit and the separate demand forecast value calculation unit, and the state of the storage battery satisfy the constraint conditions while reducing the power procurement fee for purchasing power. an electricity rate prediction unit that obtains a predicted value of the electricity rate at each time of the prediction target period;
an electricity price prediction error calculation unit that calculates a prediction error of the electricity price expected at each time of the prediction target period based on actual values of past electricity prices;
Further, a separate electricity price prediction value calculation unit calculates a separate electricity price prediction value at each time of the prediction target period using the electricity price prediction value acquired by the electricity price prediction unit and the electricity price prediction error calculated by the electricity price prediction error calculation unit,
4. The power management system according to claim 3, wherein the charge/discharge plan creation unit creates a charge/discharge plan for the storage battery by performing an optimization calculation using the plurality of electricity rate forecast values obtained by the electricity rate prediction unit and the separate electricity rate forecast value calculation unit, under constraint conditions set for the power generation forecast value, the demand forecast value, and a state of the storage battery, such that the plurality of power generation forecast values obtained by the power generation prediction unit and the separate power generation forecast value calculation unit, the plurality of demand forecast values obtained by the power demand prediction unit and the separate demand forecast value calculation unit, the plurality of consumption forecast values obtained by the vehicle state prediction unit and the separate consumption forecast value calculation unit, and a state of the storage battery satisfy the constraint conditions, while reducing the power procurement fee without causing a shortage of a storage amount required for use of the vehicle as mobility. The storage battery includes an on-board storage battery mounted in a vehicle,
a vehicle usage state prediction unit that predicts a plurality of patterns of usage states of the vehicle as a form of mobility at each time of the prediction target period and calculates an occurrence probability indicating the likelihood of each pattern occurring;
a pattern extraction unit that extracts a plurality of usage state prediction patterns from the plurality of usage state prediction patterns generated by the vehicle usage state prediction unit based on the occurrence probability;
a consumption prediction value calculation unit that calculates a prediction value of power consumption of the on-board storage battery at each time of the prediction target period for each of a plurality of usage state prediction patterns extracted by the pattern extraction unit, based on the usage state prediction pattern and an actual value of past power consumption of the on-board storage battery;
2. The power management system according to claim 1, wherein the charge/discharge plan creation unit creates a charge/discharge plan for the storage battery by performing an optimization calculation to reduce the power procurement fee without causing a shortage of the amount of storage required for use as mobility of the vehicle, while the multiple power generation forecast values obtained by the power generation forecast unit and the separate power generation forecast value calculation unit, the multiple demand forecast values obtained by the power demand forecast unit and the separate demand forecast value calculation unit, the multiple consumption forecast values obtained by the consumption forecast value calculation unit, and the state of the storage battery satisfy the constraint conditions under constraint conditions set for the power generation forecast value, the demand forecast value, and the state of the storage battery. The power management system according to any one of claims 1 to 8, characterized in that the charge/discharge plan creation unit performs an optimization calculation to reduce the power procurement fee under multiple constraint conditions set for the power generation forecast value, the demand forecast value, and the state of the storage battery, and if a solution that satisfies all of the multiple constraint conditions is not obtained, relaxes at least one of the multiple constraint conditions and performs an optimization calculation again to reduce the power procurement fee under the relaxed constraint condition. the plurality of constraint conditions include a peak value-related constraint condition for keeping a peak value of the amount of power usage due to the purchased power equal to or less than a target peak value, and a battery deterioration-related constraint condition for suppressing deterioration of the storage battery due to charging and discharging;
The power management system of claim 9, characterized in that the charge/discharge plan creation unit assigns a higher priority to the peak value-related constraint condition than to the battery degradation-related constraint condition, and repeats the optimization calculation while gradually relaxing the multiple constraint conditions in accordance with the priority until a solution that satisfies the constraint conditions is obtained. The power management system according to claim 10, characterized in that the battery degradation-related constraint condition includes a constraint condition that the stored power of the storage battery is kept above a predetermined lower limit value and below a predetermined upper limit value, and the lower limit value set for a predetermined time period within the prediction target period is set to be larger than the lower limit value set for other time periods. The power management system according to claim 10, characterized in that the battery degradation-related constraint condition includes a constraint condition that the stored power of the storage battery is kept above a predetermined lower limit value and below a predetermined upper limit value, and even if the stored power of the storage battery falls below the lower limit value, if the period during which the stored power is below the lower limit value is within a predetermined time, the current stored power is treated as the lower limit value during the predetermined time period. A charge/discharge plan creation method for creating a charge/discharge plan for a storage battery based on a predicted power generation by a power generation device, a predicted power demand by a power load, a charge/discharge amount of the storage battery, and a time-of-day electricity rate system, comprising:
A step in which a power generation prediction unit of a power management device obtains a predicted value of a power generation amount at each time of a prediction target period;
A step in which a power generation prediction error calculation unit of the power management device calculates a prediction error of the amount of power generation expected at each time of the prediction target period based on an actual value of the amount of power generation in the past;
a step of a separate power generation forecast value calculation unit of the power management device calculating a separate power generation forecast value at each time of the prediction target period using the power generation forecast value acquired by the power generation forecast unit and the power generation forecast error calculated by the power generation forecast error calculation unit;
A step in which a power demand prediction unit of the power management device obtains a predicted value of the power demand at each time of the prediction target period;
A step in which a demand prediction error calculation unit of the power management device calculates a prediction error of the amount of power demand expected at each time of the prediction target period based on an actual value of the amount of power demand in the past;
A step in which a separate demand forecast value calculation unit of the power management device calculates a separate demand forecast value at each time of the prediction target period using the demand forecast value acquired by the power demand forecasting unit and the demand forecast error calculated by the demand forecast error calculation unit;
a charge/discharge plan creation method comprising: a step of creating a charge/discharge plan for the storage battery by a charge/discharge plan creation unit of the power management device performing an optimization calculation using information on the electricity rate system for each time period during the prediction period, under constraint conditions set for the power generation forecast value, the demand forecast value, and the state of the storage battery, in which a plurality of the power generation forecast values obtained by the power generation forecast unit and the separate power generation forecast value calculation unit, a plurality of the demand forecast values obtained by the power demand forecast unit and the separate demand forecast value calculation unit, and the state of the storage battery satisfy the constraint conditions while reducing the cost of purchasing electricity.