JP2020086595A - Numerical analysis device and method - Google Patents

Abstract

To propose a numerical analysis device and a method capable of estimating analytical errors in numerical analysis and/or numerical prediction and performing effective reanalysis.SOLUTION: There are provided a numerical analysis device for performing numerical prediction and a numerical analysis method executed in the numerical analysis device. Data of an initial condition of numerical analysis and/or data of an analysis result of the numerical analysis are classified into a plurality of patterns. A relationship between each pattern and inferiority of a prediction result is detected. The inferiority of the prediction result of a numerical prediction target is estimated based on the relationship with the inferiority of the prediction result detected for the pattern in which the numerical prediction target is classified.SELECTED DRAWING: Figure 21

Description

The present invention relates to a numerical analysis device and method, and is a numerical analysis system that outputs a numerical analysis value and an estimated value of an analysis error. For example, a prediction analysis value and a prediction analysis in a numerical analysis that predicts demand and/or supply of electric power. It is suitable to be applied to a system that estimates an error and a system that estimates a weather numerical value of weather and an error of the predicted value.

The phenomenon is analyzed and the future phenomenon is predicted. For example, in the case of meteorology, by analyzing the governing equation of atmospheric motion and giving an initial value using the observed data of meteorology to the governing equation obtained by the analysis to obtain an analysis value (predicted value) of time evolution, Numerical meteorological forecasts are being conducted to predict atmospheric conditions up to several days ahead. In addition, regarding the power generation that supplies power, the power demand that fluctuates with time is analyzed and various specified times such as 1 hour ahead, 2 hours ahead, 3 hours ahead, the next day, 1 week ahead, 1 month ahead or 1 year ahead The demand for (time section) is predicted, and the operation of planning and controlling power generation is performed according to the prediction result. Regarding the power transmission system, the characteristics of the power generation unit and load unit connected to the transmission line and the power transmission unit are analyzed, and the state of the power system when the output ratio of some power generation units is changed is estimated in advance. Is being done. Similarly, in gas refining, gas demand forecasts and gas plant operation plans are being made. Furthermore, regarding air conditioning (air conditioning), by analyzing the governing equation of the motion of the air in the room and predicting the state of the air in the room, a cooler, a ventilation fan for heat exchange with the outside and intake of the atmosphere, The windows are being controlled.

Errors may occur in the analysis and/or prediction of meteorological phenomena, atmospheric phenomena, power and gas energy phenomena. Therefore, it is necessary to perform an operation that allows an error while assuming the limit of analysis, or reduce an error in analysis and/or prediction according to a condition related to the error.

In Patent Document 1, the required adjustment amount, which is the amount of power generation that should be secured in case the supply and demand forecast is missed due to changes in the weather of electric power, can be determined from the statistical value of the actual results of the weather forecast error. It is disclosed. As a result, it is possible to plan the operation of the generator according to the error of the weather forecast (meteorological analysis) for each day roughly classified by season.

Further, Patent Document 2 discloses that when an error in the meteorological analysis is found, the initial condition at the start of the analytical calculation is corrected and re-analyzed. As a result, it is possible to obtain a re-analysis value having a smaller error than the original analysis value.

JP, 2016-93106, A JP, 2008-31683, A

However, in Patent Document 1, physical characteristics of a meteorological field of the arrangement of air masses in the atmosphere (may be used as an initial condition for numerical analysis of time evolution) and characteristics of numerical fluid analysis of weather prediction (for example, The appearance of the grid data values at the end of the analysis) is not considered. Therefore, it is possible to estimate the statistically average error, but there is a problem that it is difficult to accurately estimate the error when a peculiar phenomenon that the error of the numerical analysis expands or contracts occurs.

Further, in Patent Document 2, when the observation values are continuously obtained and the analysis and reanalysis are continuously repeated, it is possible to change the initial condition based on the observation values, but the observation time when the observation data is available When there is a gap between the band and the analysis time period when the analysis value is desired, there is a problem that the analysis error cannot be detected and it is difficult to sufficiently improve the analysis accuracy.

Therefore, according to the conventional techniques disclosed in Patent Document 1 and Patent Document 2, it is not possible to sufficiently predict the analysis error that occurs or perform re-analysis to reduce the predicted analysis error. There is. There is also a problem derived from this problem that it is difficult to sufficiently optimize the operation of energy by allowing an analysis error, assuming an analysis limit.

The present invention has been made in consideration of the above points, and is intended to propose a numerical analysis apparatus and method capable of estimating an analytical error in numerical analysis and/or numerical prediction and performing effective reanalysis. ..

In order to solve such a problem, in the present invention, in a numerical analysis device that performs numerical prediction, a pattern classification unit that classifies data of initial conditions of numerical analysis and/or data of analysis results of the numerical analysis into a plurality of patterns. A detection unit that detects the relationship between each pattern and the degree of inferiority of the prediction result, and the prediction of the target of the numerical prediction based on the detection result of the detection unit for the pattern into which the target of the numerical prediction is classified. A prediction inferiority estimation unit for estimating the inferiority of the result is provided.

Further, in the present invention, there is provided a numerical analysis method executed in a numerical analysis device for performing numerical prediction, wherein data of initial conditions of numerical analysis and/or data of analysis results of the numerical analysis are classified into a plurality of patterns. Of the first step, a second step of detecting the relationship between each pattern and the inferiority of the prediction result, and the inferiority of the prediction result detected for the pattern into which the target of the numerical prediction is classified. The third step of estimating the inferiority of the prediction result of the numerical prediction target based on the relationship.

According to the present invention, it is possible to realize a numerical analysis apparatus and method capable of estimating an analysis error in numerical analysis and/or numerical prediction and performing effective reanalysis.

It is a block diagram which shows the structural example of the distributed energy operation system before energy deregulation. It is a block diagram which shows the structural example of the distributed energy operation system after energy deregulation. It is a block diagram which shows the structural example of a future distributed energy operation system. It is a block diagram which shows the structural example of a future distributed energy operation system. It is a block diagram which shows the structural example of the hardware constitutions of the energy management apparatus by 1st Embodiment. It is a block diagram which shows the structural example of a logical structure of the energy management apparatus by 1st Embodiment. It is a chart which shows the structural example of a 1st table. It is a chart showing an example of composition of a 2nd table. It is a chart showing an example of composition of a 3rd table. It is a chart showing an example of composition of a 4th table. It is a flow chart which shows a flow of a series of processings performed in an energy management device in relation to energy supply and demand. It is a flow chart which shows a processing procedure of the 1st prediction processing. It is a flow chart which shows a processing procedure of the 2nd prediction processing. It is a flow chart which shows a processing procedure of condition evaluation processing. It is a flow chart which shows a processing procedure of configuration input processing. It is a flow chart which shows a processing procedure of solution quality control processing. It is a flow chart which shows the processing procedure of the 3rd prediction processing. It is a flowchart which shows the process procedure of a planning process. It is a block diagram which shows the structural example of the logical structure of the energy management apparatus by 2nd Embodiment. It is a block diagram which shows the logical structure of the energy management apparatus by 3rd Embodiment. 9 is a flowchart for explaining the processing content of a condition evaluation unit according to the third embodiment. 9 is a flowchart for explaining the processing content of a condition evaluation unit according to the third embodiment. (A)-(G) is a figure which shows the classification example of the pattern of the columnar model of air|atmosphere. It is a figure which shows the example of classification of the pattern of the stratified state of atmospheric air. (A) is a graph used for explaining the error regression model, and (B) is a graph showing a sample distribution for each pattern. It is a flowchart which shows the flow of a series of processes performed in the energy operating device by 3rd Embodiment regarding energy supply and demand. It is a discrete diagram for explaining an error regression model.

Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings.

(1) First Embodiment (1-1) Configuration of Energy Operation System FIG. 1 shows a configuration of an energy operation system 1 (1A). The energy management system 1 is a distributed system in which a plurality of energy management apparatuses 2 (2AA, 2AB, 2AD) cooperate with each other to control energy supply and plan for it.

In FIG. 1, “EM1” indicates an energy management apparatus (EM: Energy Manager) 2 (2AA) that executes a process for controlling power interchange between a plurality of electric power companies 3, and “EM2” indicates each electric power. The energy management device 2 (2AB), “EM4”, provided at the central power supply command center of the company 3, is the energy management device 2 (2AD) of the aggregator 8 that provides the energy management service that bundles the power demands of the multiple consumers 4. ) And the energy management system 2 (2AD) installed in each customer 4 or power plant 5, respectively.

Each energy management apparatus 2 (2AB) is in a state of the energy management apparatus 2 (2AD) and 2 (2AA) which cooperate with each other (the “state” here is the power generation surplus and the power interchange between the power companies 3). Based on the power generation plan planned in advance by the own energy management system 2 (2AB) by using the remaining capacity of the system line), the current power demand status and the weather status, and the future power demand prediction and the weather prediction. A power generation command and a negative watt or positive watt generation command are transmitted to the lower energy operation device 2AD.

Then, the lowest-level energy management device 2AD (the energy management device 2AD of each customer 4 and the power plant 5) responds to the power generation command given from the higher-level energy management device 2AA, 2AB or the generation command of negative watt or positive watt. It drives the generator 6 and controls the operation of the load 7.

For example, the energy management system 2AD installed in the power station 5 drives the generator 6 in the facility according to the power generation command given from the higher-level energy management system 2AB to generate the required amount of power. Let it be done.

Further, the energy management system 2AD of the large-scale customer 4 having the power generation facility is requested by driving the generator 6 of the large-scale customer 4 according to the power generation instruction given from the higher-level energy management system 2AB. Negative watts or positive watts are generated by causing the power generation to be performed or by controlling the operation of the load 7 in the facility in accordance with the negative watt or positive watt generation command given from the higher-level energy management apparatus 2AB.

Further, the energy management system 2AD of the aggregator 8 receives the command via the energy management system 2AD installed in the contracted small-scale customer 4 in accordance with the negative or positive watt generation command given from the higher-level energy management system 2AB. By controlling the load 7 owned by the large-scale customer 4, a negative watt or a positive watt is generated.

On the other hand, in Fig. 2 in which the same parts as those in Fig. 1 are denoted by the same reference numerals, a power generation retail company that performs power generation and retailing and a regional power transmission company that performs power transmission differ under the energy liberalization system of separation of power distribution. The structural example of the energy management system 1 (1B) is shown.

Due to the liberalization of energy, the trading of electric power has been liberalized, and along with this, a trading market (hereinafter referred to as the power trading market) 2BA for trading power as a power supply has been opened. In addition, each electric power company 3 (Fig. 1) is divided into a regional power transmission company and a power generation retail company, and the energy operation device 2AB (Fig. 1) of the central power supply command station that was previously owned by the electric power company 3 is for regional power transmission. Generated power to be transferred to the regional power transmission company as the energy management device 2BBA, and to generate the power generated by the generator 6 of the large-scale customer 4 and the positive watts and the negative watts generated by the small-scale demand 4 in the electric power market 2BA. The energy management device 2BBB is newly installed by the retail company, and the energy management device 2BBB is connected to the energy management device 2AD of the large-scale customer 4 or the small-scale customer 4.

On the other hand, in recent years, it has been proposed to adjust the supply and demand across the management range of each of the plurality of electric power companies 3, and as shown in FIG. 3 in which parts corresponding to those in FIG. There may be a form of installing the energy operation device 2AC for adjusting the supply and demand. Further, as shown in FIG. 4 in which parts corresponding to those in FIG. 3 are denoted by the same reference numerals, an adjustment power trading market 2BC for trading power as adjustment power may be installed.

By the way, the energy management system 1 (1A-) in which the energy management systems 2AA to 2AD, 2BA, 2BBA, 2BBB (hereinafter collectively referred to as the energy management system 2) as described above with reference to FIGS. In 1D), the upper energy operating device 2 has a role of controlling the lower energy operating device 2, and the lowest energy operating device 2 has a role of controlling the generator 6 and the load 7.

In such a hierarchical distributed system, basically, it is necessary for the energy management system 2 to make an energy supply plan with higher accuracy as it goes down. However, the accuracy of the energy supply plan created by each energy management apparatus 2 also differs depending on the presence or absence of cooperation with another energy management apparatus 2.

Here, the “accuracy” is the degree of agreement between the calculated output value and the true value, in the sense of including accuracy (trueness) and precision (precision). The “accuracy” is the degree of the difference between the expected value and the true value of the calculated output value. "Precision" is the small variation in the independent calculation output values. In the calculation of the prediction, the fact that the calculated output value (predicted value) is accurate and the calculated output value (predicted value) is precise are not always achieved at the same time.

For this reason, it is desirable that the energy management system 2 be made more precise toward the lower rank and higher precision toward the higher rank. On the other hand, it is desirable that the energy management system has higher accuracy as it goes to the lower order, and higher accuracy as it goes to the higher order. The energy operation devices 2 having different purposes of accuracy and precision (sometimes referred to as strictness and strictness) cooperate with each other, resulting in precision (total accuracy, simply described as accuracy). In some cases), it is possible to make excellent energy forecasts and supply plans.

On the other hand, the energy demand in the management area of each energy management device 2 is determined by the weather condition in the management area and the energy demand density (traffic volume per unit area, communication generation amount per unit area, or working in all industries). Energy demand per unit area that correlates with the number of people, etc., and power generation density (the number of installed solar power generators per unit area and their respective power generation capacity, the number of installed cogeneration (electric heat supply devices) and each It depends on the power generation capacity and the energy supply and demand situation such as the number of installed generators and the power generation density that correlates with each power generation capacity), and it is necessary to plan the energy supply considering these.

Therefore, the energy management system 2 according to the present embodiment plans the energy supply in the management area of the own energy management system 2 based on its own role in the energy management system 1 and the meteorological situation and the energy supply/demand status in its management area. It is characterized by planning. Hereinafter, the energy management system 2 of this embodiment will be described.

(1-2) Configuration of Energy Operation Device of Present Embodiment FIG. 5 shows a configuration of the energy operation device 10 according to the present embodiment applied as the energy operation device 2 in FIG. The energy management apparatus 10 includes a CPU (Central Processing Unit) 11, a main memory 12, an external storage device 13, an input/output interface 14, and a plurality of communication devices 15.

The CPU 11 is a processor that controls the operation of the entire energy management apparatus 10. The main memory 12 is composed of, for example, a volatile semiconductor memory and is used as a work memory of the CPU 11.

The external storage device 13 is composed of a large-capacity nonvolatile storage device such as a hard disk device or an SSD (Solid-State Drive), and is used to store various programs and data for a long period of time. Each program that configures a reference data storage unit 30, first to third prediction units 31 to 33, a condition evaluation unit 34, a configuration input unit 35, a solution quality control unit 36, and a planning unit 37, which will be described later with reference to FIG. It is stored and held in the external storage device 13.

The program stored in the external storage device 13 is loaded into the main memory 12 when the energy management apparatus 10 is started or when necessary, and the CPU 11 executes the loaded program, whereby the energy management apparatus 10 as a whole is described below. Various processes are performed.

The input/output interface 14 includes an external communication terminal 20, an input device 21, and a display device 22. The external communication terminal 20 is composed of, for example, a USB (Universal Serial Bus) terminal. The input device 21 is composed of a keyboard, a mouse, and/or a touch panel, and is used by the user to input necessary information. The display device 22 is composed of, for example, a liquid crystal panel or an organic EL (Electro Luminescence) panel, and is used to display necessary screens and information.

The communication device 15 is a server device, load facility, and/or power generation facility of a government agency, a company or an organization, in which the energy management system 10 provides another energy management system 10, energy information services such as weather information and traffic information. Controls the protocol when communicating with.

FIG. 6 shows a logical configuration of the present energy management apparatus 10. As shown in FIG. 6, the energy management apparatus 10 includes a reference data storage unit 30, a first prediction unit 31, a second prediction unit 32, a third prediction unit 33, a condition evaluation unit 34, a configuration input unit 35, and a solution. It comprises a quality control unit 36 and a planning unit 37.

The reference data storage unit 30 is configured by a partial storage area of the main memory 12 or the external storage device 13, and includes a demand data storage unit 40, a weather data storage unit 41, a demand factor data storage unit 42, and a market data storage unit 43. Prepare

The demand data storage unit 40 is a database that stores past energy demand in the management area of the self-energy operating apparatus 10. The demand data storage unit 40 sequentially stores, as demand data, the power consumption amounts of these consumers, which are periodically acquired from the power meter systems 44 of the consumers in the management area.

The meteorological data storage unit 41 is a database that stores past observation data (weather data) of each meteorological element such as temperature, solar radiation, humidity, wind speed, wind direction, and atmospheric pressure. In the meteorological data storage unit 41, the observation data of each meteorological element for each meteorological grid, which is periodically acquired from the meteorological observation system 45 of the Meteorological Agency or a private weather forecasting company, and the weather forecasting system 46 of the meteorological agency or a private weather forecasting company. The forecast data (meteorological forecast data) of each weather element for each weather grid, which is periodically obtained from, is sequentially stored. Note that the “weather grid” refers to points that are set at regular intervals in the latitude direction, the longitude direction, and the vertical direction on the ground surface and in the space from the ground surface to the sky above a predetermined height.

The demand factor data storage unit 42 includes various factors of electric power demand such as road traffic volume, communication generation amount, number of industrial workers, and number of railway passengers in the management area of the energy management apparatus 10 (hereinafter, these are referred to as demand factors). ) Is a database that stores observed values. In the demand factor data storage unit 42, for example, an observation value regarding the road traffic volume in the management area of the self-energy operation apparatus 10 which is periodically acquired from the traffic data distribution server 47 of the road traffic public corporation, and data of each large consumer. Observed values related to each demand factor such as observed values of the communication generation amount of the large-scale consumer, which are periodically acquired from the distribution server 48, are sequentially stored.

The market data storage unit 43 is a database that stores information on transactions in the electricity trading market. The market data storage unit 43 stores, as market data, information such as power bidding information and unit price in the power trading market.

On the other hand, the first prediction unit 31 is a functional unit having a function of predicting the weather in the management area of the self-energy management apparatus 10 with the quality specified by the solution quality control unit 36 as described later. The first prediction unit 31 includes a first data assimilation unit 52 including a reference point adjustment processing unit 50 and a data assimilation processing unit 51, a mobile observation reference point adjustment processing unit 53, a sequential data assimilation processing unit 54, and a physical unit. And a simulation unit 55.

The reference point adjustment processing unit 50 of the first data assimilation unit 52 is embodied by the CPU 11 executing the corresponding reference point adjustment processing program 50P (FIG. 5) stored in the external storage device 13 (FIG. 5). It is a functional part. In order to predict the weather in the management area with the quality designated by the solution quality control unit 36, the reference point adjustment processing unit 50 should emphasize the weather in the data assimilation process executed by the data assimilation processing unit 51 as described later. Determine the grid as a reference point.

As a matter of fact, the first predicting unit 31 uses a numerical value of 1 to 5 as the quality, which will be described later, indicating the stability of the prediction result that the energy management apparatus 10 should target (hereinafter, this is the predicted solution target stability). A solution quality control unit 36 notifies the solution. In addition, as shown in FIG. 7, the first predicting unit 31 predefines a first table 56 in which the user determines in advance which weather grid should be determined as a reference point for each value of the predicted solution target stability index. Holding

Thus, the reference point adjustment processing unit 50 sets the weather grid defined in the first table 56 for the value of the predicted solution target stability index given by the solution quality control unit 36 (in FIG. A certain "upper weather grid" or "surface weather grid" that is each weather grid on the ground surface), and determines the weather grid in the management area of the self-energy operating apparatus 10 as the reference point, and assimilates the data. Notify the processing unit 51.

In general, when weather forecasts are performed using meteorological data obtained near the ground surface, high-precision forecasting results are obtained because the weather data directly measured on the ground surface are used for forecasting. Since the solar radiation, atmospheric pressure, wind direction, air volume, etc. are unstable near the surface, the prediction result is also unstable (the hitting is large). On the other hand, when performing meteorological forecasts using the meteorological data acquired in the sky, stable forecast results (prediction results with few deviations) are obtained because the temperature, pressure, wind direction, and air volume are stable in the sky. However, only the prediction result with low accuracy is obtained. Therefore, in the first table 56, the higher the value of the predicted solution target stability index (the higher the quality of stability) is associated with the upper layer weather grid in which the weather condition is stable, and the value of the predicted solution target stability index is The smaller the value (the higher the precision, in spite of the lower quality of the stability of the solution, the higher the accuracy) is associated with the meteorological grid near the ground surface, which gives a highly accurate prediction result.

Further, the data assimilation processing unit 51 of the first data assimilation unit 52 is a function implemented by the CPU 11 (FIG. 5) executing the corresponding data assimilation processing program 51P (FIG. 5) stored in the external storage device 13. It is a department. The data assimilation processing unit 51 stores observation data (weather data) that emphasizes the reference points determined by the reference point adjustment processing unit 50, and simulation result data of the physical simulation executed by the physical simulation unit 55 as described later. Execute the process of assimilating data.

In practice, the data assimilation processing unit 51 reads out the meteorological data of each of the above-mentioned reference points from the meteorological data storage unit 41, and uses the meteorological data and the simulation result of the physical simulation unit 55 by the reference point adjustment processing unit 50. A data assimilation process is performed in which data assimilation is performed while emphasizing the determined reference points. As a result, the past can be reproduced from such meteorological data to estimate (predict) the value in the future time section.

The mobile observation reference point adjustment processing unit 53 is implemented by the CPU 11 executing a corresponding mobile observation reference point adjustment processing program 53P (FIG. 5) stored in the external storage device 13 (FIG. 5). It is a functional unit. The mobile observation reference point adjustment processing unit 53 is a simulation model generated by the physical simulation unit 55 as described later based on the current weather data obtained by mobile observation using a weather balloon or an observation radar. Determine a reference point to correct for. Specifically, the mobile observation reference point adjustment processing unit 53, in the same manner as the reference point adjustment processing unit 50, sequentially performs the weather forecast in the management area with the quality specified by the solution quality control unit 36. Reference points to be emphasized in the data assimilation process executed by the data assimilation processing unit 54 are determined from the weather grid in which the current weather data is obtained.

The sequential data assimilation processing unit 54 is a functional unit implemented by the CPU 11 executing the sequential data assimilation processing program 54P (FIG. 5) stored in the external storage device 13. The sequential data assimilation processing unit 54 dynamically adjusts the meteorological data of each meteorological grid determined as the reference points by the mobile observation reference point adjustment processing unit 53 and the data obtained by the physical simulation of the physical simulation unit 55. A data assimilation process is performed for assimilating data while emphasizing the reference points determined by the observation reference point adjustment processing unit 53.

The physical simulation unit 55 is a functional unit realized by the CPU 11 executing the physical simulation program 55P (FIG. 5) stored in the external storage device 13. The physical simulation unit 55 is based on the weather data of each reference point given from the data assimilation processing unit 51 of the first data assimilation unit 52 and the current weather data of each reference point given from the sequential data assimilation processing unit 54. , Perform a physical simulation (numerical simulation) of each meteorological element at each of these reference points.

Specifically, the physics simulation unit 55 follows the thermodynamic equation of the atmosphere consisting of gas and water vapor, and the equations of motion of the mass and energy conservation of the atmosphere, and changes in the atmospheric temperature and the amount of water vapor at every minute time and the associated surface solar radiation. The value of each meteorological element such as is estimated by the simulation of the meteorological system in which the radiant energy from the sun is input. Then, the physical simulation unit 55 outputs the simulation result of the physical simulation to the condition evaluation unit 34.

In the first prediction unit 31 having such a configuration, for example, when the predicted solution target stability index given from the solution quality control unit 36 is “4”, the reference point adjustment processing unit 50 refers to it as described above. The "upper weather grid" is decided as a point, and weather forecast is performed based on the meteorological data of these "upper weather grid" acquired by aircraft observation, radar observation, balloon observation, etc. For the meteorological grid of which one is missing, the meteorological data observed on the ground of the surrounding meteorological grid shall be substituted.

The substitution is achieved by using an observer in the Kalman filter that estimates the value of the upper weather grid from the meteorological data observed on the ground, or by using the value estimated by the physical simulation of the local weather grid. By doing so, stable prediction can be performed by a physical simulation in an upper layer that is physically stable, and a relatively accurate approximation can be obtained even when the meteorological phenomenon is unstable. This is because, when performing a physical simulation of weather in a large spatial area with a large time granularity, the physical simulation with good accuracy can be performed by using the meteorological data of the upper meteorological grid.

Further, in the first prediction unit 31, for example, when the predicted solution target stability index given from the solution quality control unit 36 is "1", the meteorological data of the "surface weather grid" will be used. However, due to this, a short time interval of less than 30 minutes (fine time granularity), and detailed observation of the management area of the self-energy operating apparatus 10 (fine spatial granularity) and highly precise prediction can be made by physical simulation. It is possible to obtain a more accurate forecast value when the meteorological phenomenon is stable.

The second prediction unit 32 specifies the quality specified by the solution quality control unit 36 as described later (prediction stability according to the value of the predicted solution target stability index given by the solution quality control unit 36 as described below). And a second data assimilation unit 62 including a reference point adjustment processing unit 60 and a data assimilation processing unit 61, which is a functional unit having a function of predicting the data demand density in the management area of the self-energy operation device 10. And a physical simulation section 63.

The reference point adjustment processing unit 60 of the second data assimilation unit 62 is a functional unit realized by the CPU 11 executing the corresponding reference point adjustment processing program 60P (FIG. 5) stored in the external storage device 13. .. The reference point adjustment processing unit 60 is described later in order to predict the energy demand density in the management area with the quality (stability of prediction) according to the value of the predicted solution target stability index notified from the solution quality control unit 36. As described above, the reference points to be emphasized in the data assimilation processing executed by the data assimilation processing unit 61 are determined.

Specifically, the reference point adjustment processing unit 60 has a larger temporal granularity and spatial granularity as the value of the predicted solution target stability index notified from the solution quality control unit 36 is larger, and has a smaller value of the predicted solution target stability index. Several points on the ground surface within the management area of the self-energy operating apparatus 10 are determined as reference points so that the temporal granularity and the spatial granularity become smaller. Then, the reference point adjustment processing unit 60 notifies the data assimilation processing unit 61 of the point determined as the reference point.

The data assimilation processing unit 61 is a functional unit implemented by the CPU 11 executing the corresponding data assimilation processing program 61P (FIG. 5) stored in the external storage device 13. The data assimilation processing unit 61, similar to the data assimilation processing unit 51 of the first prediction unit 31, the energy demand density data based on the observation data of the corresponding demand factors accumulated in the demand factor data accumulation unit 42, and As described above, the data assimilation process for assimilating the simulation result data of the physical simulation executed by the physical simulation unit 63 with the reference points determined by the reference point adjustment processing unit 60 is emphasized.

In practice, the data assimilation processing unit 61 reads out the observed values of each demand factor at each of the above-mentioned points determined by the reference point adjustment processing unit 60 from the demand factor data storage unit 42, and the observed values of each of these demand factors, Using the simulation result of the physical simulation for each demand factor executed in the physical simulation unit 63, the past is reproduced from the observed value of each demand factor to estimate (predict) the value in the future time section. Perform data assimilation processing.

The physical simulation unit 63 is a functional unit implemented by the CPU 11 executing the physical simulation program 63P (FIG. 1) stored in the external storage device 13. The physics simulation unit 63, based on the data assimilation-processed observation value of each demand factor at each reference point provided from the data assimilation processing unit 61, performs a numerical simulation of the energy demand density at each of these reference points for each demand factor ( A multi-agent simulation or a simulation simulating a demand factor as a fluid is executed. Then, the physical simulation unit 63 outputs the simulation result of the numerical simulation to the third prediction unit 33 and the condition evaluation unit 34.

The third predicting unit 33, the demand data accumulated in the demand data accumulating unit 40, the meteorological data accumulated in the meteorological data accumulating unit 41, the observation data of each demand factor accumulated in the demand factor data accumulating unit 42, and Based on the necessary data of the market data accumulated in the market data accumulation unit 43 and the energy demand density in the management area predicted by the second prediction unit 32, the solution quality control unit 36 Predict the energy demand in the management area of the self-energy management apparatus 10 with the designated quality (stability of prediction according to the value of the predicted solution target stability index given from the solution quality control unit 36 as described later). It is a functional unit having a function and includes a reference point adjustment processing unit 70, a regression model generation unit 71, and a regression prediction processing unit 72.

The reference point adjustment processing unit 70 is a functional unit realized by the CPU 11 executing the corresponding reference point adjustment processing program 70P (FIG. 5) stored in the external storage device 13. The reference point adjustment processing unit 70 predicts the energy demand in the management area of the self-energy management apparatus 10 with the quality according to the value of the predicted solution target stability index designated by the solution quality control unit 36 at that time. The regression model generation unit 71 selects the type of regression model to be generated and the conditions of the regression model.

In practice, the reference point adjustment processing unit 70 uses the predicted solution target stability index specified by the solution quality control unit 36 for the type of regression model to be generated by the regression model generation unit 71 and its regression model. A second table 73 as shown in FIG. 8 which defines conditions of explanatory variables (each weather element, each demand factor, and/or data demand density) is held in advance. The second table 73 is created in advance by the user. In this case, a threshold value of the importance level P of the explanatory variable (“threshold” column in FIG. 8) is set as the “condition of the explanatory variable used in the regression model”.

Thus, the reference point adjustment processing unit 70 sets the regression model defined in the second table 73 for the value of the predicted solution target stability index specified by the solution quality control unit 36 and the explanatory variables used for the regression model. The conditions (threshold value of the importance of the explanatory variable) are selected and output to the regression model generation unit 71.

The regression model generation unit 71, based on the type of the regression model given from the reference point adjustment processing unit 70 and the conditions of the explanatory variables used for the regression model, describes the energy demand and the description in the management area of the self-energy management apparatus 10. A regression model with the variables is generated, and the generated regression model is output to the regression prediction processing unit 72.

Specifically, the regression model generation unit 71 calculates an index of the degree of importance P for each weather element, each demand factor, and each data demand density that can be an explanatory variable by using a generally known random forest algorithm, and Each meteorological element, each demand factor, and each data that normalizes the index to 0 to 1 and the value of the normalized importance P that satisfies the “condition of the explanatory variable used for the regression model” notified from the reference point adjustment processing unit 70 The regression model notified from the reference point adjustment processing unit 70 is generated using the demand density as an explanatory variable.

In addition, if there are explanatory variables for each point of a weather station (for example, Tokyo, Maebashi, Yokohama) or for each unit plane that manages demand (for example, for each weather grid), multiple explanatory variables can be reduced. Data of points (or unit planes) may be aggregated into one data which is a weighted average. When the weighted average is calculated, the demand density and the power generation density calculated by the second prediction unit are used as the weighting coefficient. By aggregating the data in this way, the prediction results will be more average.

Therefore, for example, when the value of the predicted solution target stability index given from the solution quality control unit 36 is “5”, the weather factor, the demand factor, and/or the normalized importance P of “0.9” or more. A "Lasso regression model" with the data demand density as an explanatory variable will be generated. By limiting the explanatory variables of the regression model in this way, it is possible to appropriately simplify the generated regression model. By using a simplified regression model with few explanatory variables, stable energy demand can be predicted.

Further, when the value of the predicted solution target stability index given from the solution quality control unit 36 is “1”, the meteorological element, the demand factor, and the data demand density with the normalized importance P of “0.2” or more are calculated. A strict “Ridge regression model” with many explanatory variables is generated as an explanatory variable. As a result, it is possible to accurately predict the energy demand, and it is possible to obtain a highly accurate predicted value particularly when the weather and demand factors are within the normal range and the prediction by the regression model is easy.

As described above, in the second table 73, the larger the value of the predicted solution target stability index (the higher the quality), the simpler the regression model is designated as the regression model, and the threshold value of the importance P of the explanatory variable. Is set to a larger value, and the smaller the value of the predicted solution stability index is (the lower the quality is), the more complicated the regression model is set as the regression model, and the smaller the threshold value of the importance P of the explanatory variable is set. .

The regression prediction processing unit 72 is a functional unit implemented by the CPU 11 executing the corresponding regression prediction processing program 72P (FIG. 5) stored in the external storage device 13. The regression prediction processing unit 72 acquires the value of the necessary explanatory variable from the simulation result provided from the meteorological data storage unit 41, the demand factor data storage unit 42, and the physical simulation unit 63 of the second prediction unit 32, and the acquired value is acquired. The predicted value of the energy demand amount of the time section (designated time) in the management area of the self-energy operating apparatus 10 is calculated by applying the regression model given from the regression model generation unit 71. Then, the regression prediction processing unit 72 outputs the calculated predicted value of energy demand to the planning processing unit 101 of the planning unit 37 described later.

Note that instead of the regression prediction processing executed by the regression prediction processing unit 72 of the present embodiment, time series prediction, similar day prediction using past demand records of similar calendar days and meteorological conditions, and similar day candidates. Demand pattern prediction may be performed using a pattern of demand changes during a predetermined period such as one day/one week. For example, in time series prediction, when stable prediction is performed, it is possible to perform prediction with a wide sampling interval and ignore minute fluctuations (fringes) in demand. For similar day prediction, increase or decrease the number of occasions for holidays and weather conditions on calendar days, and reduce the number of occasions for stable forecasting, and use the average value of past results as the forecast value. be able to.

On the other hand, the condition evaluation unit 34 determines the distribution shape of data for each time section and space section for each simulation result of the physical simulation executed by each physical simulation section 55, 63 of the first and second prediction sections 31, 32. The functional unit has a function of evaluating each, and includes a temporal data distribution shape evaluation unit 80, a spatial data distribution shape evaluation unit 81, and a tag data writing unit 82.

The temporal data distribution shape evaluation unit 80 is a functional unit implemented by the CPU 11 executing the temporal data distribution shape evaluation program 80P (FIG. 5) stored in the external storage device 13. The temporal data distribution shape evaluation unit 80 determines the magnitude of the temporal change in the prediction result of the meteorological data of each reference point obtained as the simulation result of the physical simulation executed by the physical simulation unit 55 of the first prediction unit 31. For example, the weighted average is evaluated by the dispersion state (dispersion value) of these predicted values.

The spatial data distribution shape evaluation unit 81 is a functional unit realized by the CPU 11 executing the spatial data distribution shape evaluation program 81P (FIG. 5) stored in the external storage device 13. The spatial data distribution shape evaluation unit 81 calculates the horizontal direction based on the prediction result of each reference point obtained by the physical simulation executed by each physical simulation unit 55, 63 of the first and second prediction units 31, 32. Evaluate the change in the predicted value between each reference point on the same plane parallel to and the reference point on the same plane parallel to the vertical direction by the variance of the predicted values (variance value) in the same time section ..

The tag data writing unit 82 is a functional unit realized by the CPU 11 executing the tag data writing program 82P (FIG. 5) stored in the external storage device 13. The tag data write-out unit 82 of the first prediction unit 31 is based on the evaluation result (dispersion value) of the temporal data distribution shape evaluation unit 80 and the evaluation result (dispersion value) of the spatial data distribution shape evaluation unit 81. Future weather conditions (hereinafter, referred to as weather conditions) obtained by the physical simulation executed by the physical simulation unit 55 and the simulation executed by the physical simulation unit 63 of the second prediction unit 32. Stability with the future energy demand density situation (hereinafter referred to as energy demand density condition) is determined.

Practically, the tag data writing unit 82 sets the evaluation result (dispersion value) of the temporal data distribution shape evaluation unit 80 and the evaluation result (dispersion value) of the spatial data distribution shape evaluation unit 81 to preset threshold values. It is determined whether the future weather condition or energy demand density condition is stable (small dispersion) or unstable (large dispersion) by comparing with the above, and the determination result is stored in, for example, the external storage device 13. In a predetermined storage area of.

Note that in the case of the present embodiment, the tag value indicating the stability of the weather condition includes “small disturbance” indicating stability and “large disturbance” indicating instability, which indicates stability of the energy demand density condition. The tag value includes “moderate” indicating stability, “severe” indicating instability, and “normal” indicating an intermediate state between stable and unstable.

The configuration input unit 35 is a functional unit implemented by the CPU 11 executing the configuration input program 35P (FIG. 5) stored in the external storage device 13. The configuration input unit 35 is a target stabilization for a lower EM, which will be described later, which is notified from the higher-level energy management apparatus 10 in cooperation or set by the user using the input device 21 (FIG. 5) of the input/output interface 14 (FIG. 5). The index is fetched and the fetched target stability index for lower EM is output to the solution quality control unit 36.

The solution quality control unit 36 is a simulation (physical simulation or simulation of a predicted value by a regression model) in the first to third prediction units 31 to 33, and a simulation executed in a planning unit 37 described later (plans such as interior points). This is a functional unit having a function of controlling the quality of the solution of the approximate solution of the problem and the simulation of the control calculation.

The solution quality control unit 36 receives from the configuration input unit 35 the tag value of the weather condition and the tag value of the data demand density condition which the tag data writing unit 82 of the condition evaluation unit 34 has written in a predetermined area in the external storage device 13. Based on the given target stability index for lower EM and the third table 90 shown in FIG. 9, the relaxation solution target stability index described later and the predicted solution target stability for the above-described first to third prediction units 31 to 33. The index and the above-mentioned target stability index for lower-order EM to be given to the lower-level energy management apparatus 10 are respectively acquired, and these acquired target stability indexes are set in the planning unit 37, the first to third prediction units 31 to 33, and the lower level. Output to the energy management device 10.

Here, the third table 90 will be described. The third table 90 is provided with the above-described mitigation solution target that is preset for each combination pattern of the tag value of the weather condition and the tag value of the energy demand density condition (hereinafter, these are referred to as condition tag value combination pattern). It is a table used to manage the stability index, each predicted solution target stability index to the first to third prediction units 31 to 33, and the target stability index for the lower EM, and is dedicated by the user in advance for each energy management device 10. Is created.

As shown in FIG. 9, the third table 90 includes a role tag column 90A, a weather condition evaluation tag value column 90B, an energy demand density condition evaluation tag value column 90C, a first predicted solution target stability index column 90D, and a second table. The prediction solution target stability index column 90E, the third prediction solution target stability index column 90F, the relaxation solution target stability index column 90G, and the target stability index column 90H for lower EM are configured. In the third table 90, one row corresponds to one condition tag value combination pattern, and the same number of rows as the number of all condition tag value combination patterns are provided.

The role of the energy management apparatus 10 in the entire energy management system 1 is stored in the role tag column 90A. In the weather condition evaluation tag value column 90B and the energy demand density condition evaluation tag value column 90C, the tag value of the weather condition and the tag value of the energy demand density condition in the corresponding condition tag value combination pattern are stored.

Further, in the first predicted solution target stability index column 90D, the second predicted solution target stability index column 90E, and the third predicted solution target stability index column 90F, the first preset values for the corresponding condition tag value combination patterns are set. ~ The predicted solution target stability indices for the third prediction units 31 to 33 are stored. Furthermore, the relaxation solution target stability index column 90G stores the relaxation solution target stability index preset for the corresponding condition tag value combination pattern, and the target stability index column 90H for lower EMs stores the corresponding condition tag value. A target stability index for the lower EM preset for the combination pattern is stored.

Therefore, in the case of the example of FIG. 9, for example, regarding the condition tag value combination pattern in which the tag value of the weather condition is “disturbance large” and the tag value of the energy demand density condition is “normal”, the first to third prediction units The predicted solution target stability indices for 31 to 33 are set to “4”, “3”, and “4”, respectively, the relaxation solution target stability index is set to “5”, and the target stability index for lower EM is set to “4”. It has been set.

Note that in the third table 90, for a more stable condition tag value combination pattern, the relaxation solution target stability index, each predicted solution target stability index, and the target stability index for lower EMs are set so that stable energy demand can be predicted. The value is also set to a large value overall, and for more unstable condition tag value combination patterns, the relaxation solution target stability index, each predicted solution target stability index, and the lower EM so that more accurate energy demand prediction can be performed. The target stability index is also set to a small value overall.

In addition, in the energy management system 1 of FIG. 5, the third table of the energy management system 10 that is not linked to the lower-level energy management system 10 and other energy management systems 10 (that is, there is no energy management system 10 that backs up). 90 is an overall relaxation solution target stability index and each predicted solution target so that more accurate prediction can be made as compared with the third table 90 of the higher-level energy management apparatus 10 and other energy management apparatus 10. The values of the stability index and the target stability index for lower EM are set to be large.

In addition to the configuration in which the target stability index is notified from the energy management apparatus 10 that is linked as in the present embodiment, the energy that the self-energy management apparatus links to the remaining amount of power and adjustment power that is a margin in planning and control. The operation device 10 may be notified. The energy management apparatus 10 that has received the notification sets the values of the relaxation solution target stability index, each predicted solution target stability index, and the target stability index for the lower EMs when the other energy management apparatus serving as a backup has a small margin. The third table 90 is automatically updated so as to be large, or the user appropriately updates the third table 90 of the energy management apparatus 10 as described above. In this way, in the situation where there is no backup, the energy management apparatus 10 is defined as having the ultimate responsibility of its own role, and even if there is some fluctuation in demand and weather conditions, the energy supply is not affected. Thus, high-quality (high accuracy, low precision) prediction and planning of stability with margin can be performed.

On the other hand, the solution quality control unit 36 includes a relaxation solution target stability index calculation unit 91, a predicted solution target stability index calculation unit 92, and a target EM target stability index calculation unit 93 for lower EMs.

The relaxation solution target stability index calculation unit 91 is a functional unit realized by the CPU 11 executing the relaxation solution target stability index calculation program 91P (FIG. 5) stored in the external storage device 13. The relaxation solution target stability index calculation unit 91 calculates the third value based on each tag value of the weather condition and the energy demand density condition written by the tag data writing unit 82 of the condition evaluation unit 34 in a predetermined storage area of the external storage device 13. The numerical value stored in the relaxation solution target stability index column 90G of the row corresponding to the combination pattern of these tag values (condition tag value combination pattern) in the table 90 is read out to the planning unit 37 as the relaxation solution target stability index. Output.

The predicted solution target stability index calculation unit 92 is a functional unit implemented by the CPU 11 executing the predicted solution target stability index calculation program 92P (FIG. 5) stored in the external storage device 13. The predictive solution target stability index calculation unit 92, based on the tag values of the weather condition and the energy demand density condition, calculates the row of the row corresponding to the combination pattern of these tag values (condition tag value combination pattern) in the third table 90. The numerical values stored in the first predicted solution target stability index column 90D, the second predicted solution target stability index column 90E, and the third predicted solution target stability index column 90F are read out. Further, the predicted solution target stability index calculation unit 92 uses the numerical values stored in the first predicted solution target stability index column 90D among these read values as the first prediction unit 31 and the second predicted solution target stability index column 90E. The numerical value stored in the second prediction unit 32 and the numerical value stored in the third predicted solution target stability index column 90F are output to the third prediction unit 33 as the predicted solution target stability index, respectively.

Further, the target stability index calculation unit 93 for the lower EM is a functional unit realized by the CPU 11 executing the target stability index calculation program 93P (FIG. 5) for the lower EM stored in the external storage device 13. The target stability index calculation unit 93 for lower EMs, based on each tag value of the weather condition and the energy demand density condition, a row corresponding to a combination pattern of these tag values (condition tag value combination pattern) in the third table 90. The numerical value stored in the target stability index column 90H for the lower EM is read and transmitted as the target stability index for the lower EM to the lower energy operation device 10 via the corresponding communication device 15 (FIG. 5).

The relaxation solution target stability index calculation unit 91, the predicted solution target stability index calculation unit 92, and the lower EM target stability index calculation unit 93 relax based on the lower EM target stability index given from the configuration input unit 35. The solution target stability index, the predicted solution target stability index, or the target stability index for the lower EM may be corrected. In this case, the relaxation solution target stability index, the predicted solution target stability index, or the target stability index for the lower EM is A, and the target stability index for the lower EM notified from the upper energy management device is B (for example, “0.2”). , The constant is a (eg, “3”), the corrected relaxation solution target stability index, the predicted solution target stability index or the target stability index for lower EM is A′, and the relaxation solution target stability index calculation unit 91 and the predicted solution. The target stability index calculation unit 92 and the target stability index calculation unit 93 for the lower EM are
The corrected relaxation target stability index, predicted solution target stability index, or target EM-target stability index after correction may be calculated by.

However, when corrected in this way, the relaxed solution target stability index, the predicted solution target stability index, and/or the target EM target stability index after correction may have values below the decimal point. In this case, the relaxed solution target stability index calculation unit 91, the predicted solution target stability index calculation unit 92, and the target stability index calculation unit 93 for the lower EM calculate the relaxed solution target stability index or the predicted solution target stability index or The target stability index for the lower EM is output as it is to the planning unit 37, the first to third prediction units 31 to 33 or the lower energy operation device 10, and the relaxation solution target stability index, the predicted solution target stability index, or the lower EM is output. In the planning unit 37, the first to third prediction units 31 to 33, or the lower-level energy management device 10 that has received the target stability index, the relaxation solution target stability index, the predicted solution target stability index, or the integer of the target stability index for the lower EM. The energy management apparatus 10 may be constructed so as to execute the same processing as described above based on the value of the part.

On the other hand, the planning unit 37 predicts the energy demand given from the regression prediction processing unit 72 of the third prediction unit 33, and the relaxation solution target given from the relaxation solution target stability index calculation unit 91 of the solution quality control unit 36. It is a functional unit having a function of making a plan according to the role of the self-energy operating apparatus 10 based on the stability index, and includes a constraint relaxation processing unit 100 and a planning processing unit 101.

The constraint relaxation processing unit 100 is a functional unit implemented by the CPU 11 executing the constraint relaxation processing program 100P (FIG. 5) stored in the external storage device 13. The constraint relaxation processing unit 100, based on the value of the relaxation solution target stability index given by the relaxation solution target stability index calculation unit 91 of the solution quality control unit 36, is an exact solution to the plan created by the plan processing unit 101 as described later. The permissible gap ratio is set in the plan processing unit 101.

In practice, the constraint relaxation processing unit 100 holds a fourth table 102 in which the allowable gap ratio from the exact solution set in advance for each value of the relaxation solution target stability index is defined, as shown in FIG. is doing. The fourth table 102 is created in advance by the user. In the fourth table 102, the larger the value of the predicted solution target stability index (the higher the quality), the larger the allowable gap ratio so that the prediction result can be obtained more stably and quickly. The smaller the value of (the lower the quality), the smaller the allowable gap ratio is set so that a more accurate prediction result can be obtained.

Then, when the relaxation solution target stability index calculation unit 91 of the solution quality control unit 36 provides the relaxation solution target stability index, the constraint relaxation processing unit 100 determines the allowable gap ratio set for the value of the relaxation solution target stability index. 4 is read from the table 102, and the read allowable gap ratio is set in the plan processing unit 101.

The plan processing unit 101 is a functional unit realized by the CPU 11 executing the plan processing program 101P (FIG. 5) stored in the external storage device 13. The plan processing unit 101, based on the predicted value of the energy demand given from the regression prediction processing unit 72 of the third prediction unit 33, within the range of the allowable gap ratio set by the constraint relaxation processing unit 100, the own energy operation device 10 It plans the energy supply according to the role of, and controls the corresponding external equipment according to the plan.

For example, in the energy management system 1 shown in FIG. 4, the energy management system 2AD of each customer 4 and the energy management system 2AD of the power plant 5 control the generator 6 and the load 7 connected to the own energy management system 2AD. A plan is prepared, and the generator 6 and the load 7 are controlled according to the prepared plan. In addition, the energy management system 2AD of the aggregator 8 makes a plan for controlling the lower energy management system 2AD connected to the own energy management system 2AD, and controls these lower energy management systems 2AD according to the planned plan.

Further, the energy management device 2BBB of the energy management company of the power generation retailer makes a plan for receiving and ordering power from the power exchange and a plan for controlling the generator 6 and the load 6 in the management area, and according to the plan, It receives and orders electric power from the exchange and gives an instruction regarding energy supply to the lower energy operation device 2AD. Normally, the power exchange 2BA conducts power transactions in units of 30 minutes (or in units of 15 minutes or 1 hour), so the energy management system 2BBB also makes predictions and plans in units of 30 minutes.

Further, the energy management system 2BBA of the regional power transmission company plans and controls the generation of instantaneous electric power (generation of positive watt and negative watt) including the time interval of less than 30 minutes to coincide with the instantaneous electric power consumption. .. Instantaneous generation of electric power instantaneously according to fluctuations in actual demand is called adjustment power, and the energy management system 2BBA predicts the amount of adjustment power required for future time sections and adjusts the exchange power. A plan for ordering the adjustment power to the 2BC is made, and a control instruction is given to the energy operation device 2AD of the power plant that is the supplier of the ordered adjustment power and the energy operation device 2AD of the aggregator 8. The energy management system 2AD of the power plant and the energy management system 2AD of the aggregator 8 follow the command of the adjustment power control with priority over the others.

(1-3) Process Flow (1-3-1) Process Flow in Energy Operation Device FIG. 11 shows a series of process flows related to energy supply planning and control executed in the energy operation device 10 (hereinafter, This is called an energy supply plan and control process). This energy supply plan and control process are executed regularly.

In the energy management system 10, first, in the solution quality control unit 36 (FIG. 6), the tag data writing unit 82 (FIG. 6) of the condition evaluation unit 34 (FIG. 6) is used for external storage in the previous energy supply plan and control process. The tag value of the weather condition and the tag value of the energy demand density condition written in a predetermined storage area of the device 13 and the lower order given from the configuration input unit 35 (FIG. 6) in the previous energy supply plan and control process. Based on the EM target stability index, the relaxation solution target stability index, the predicted solution target stability index given to the first to third prediction units 31 to 33 (FIG. 6), and the lower EM given to the lower energy operation device 10. And a target stability index for each of them are calculated (S1).

Subsequently, the first prediction unit 31 calculates the self-energy based on the past weather data accumulated in the weather data accumulation unit 41 (FIG. 6) and the predicted solution target stability index given by the solution quality control unit 36. The future weather in the management area of the operation device 10 is predicted, and the prediction result is output to the condition evaluation unit 34 (S2).

Then, the second predicting unit 32 operates its own energy based on the observed value of each demand factor accumulated in the demand factor data accumulating unit 42 and the predicted solution target stability index given from the solution quality control unit 36. The future energy demand density in the management area of the device 10 is predicted, and the prediction results are output to the third prediction unit 33 and the condition evaluation unit 34 (S3).

Then, the condition evaluation unit 34 manages the weather forecast result of the weather in the management area of the self-energy management device 10 provided by the first prediction unit 31 and the self-energy management device 10 provided by the second prediction unit 32. Based on the forecast result of the future energy demand density in the area, the temporal and spatial distribution of the weather condition and the energy demand density condition is evaluated, and the tag value indicating the future condition of the weather condition and the energy demand density The tag value indicating the future condition of the condition and the tag value are written to the predetermined storage areas of the external storage device 13 (S4).

Subsequently, the configuration input unit 35 (FIG. 6) is given from the upper energy operation apparatus 10 or the lower EM specified by the user using the input device 21 (FIG. 5) of the input/output interface 14 (FIG. 5). The target target stability index is fetched and output to the solution quality control unit 36 (S5).

Next, the solution quality control unit 36 gives the tag values of the weather condition and the energy demand density condition which the condition evaluation unit 34 has written to the predetermined storage area of the external storage device 13 at that time, and the configuration input unit 35 at this time. Based on the obtained target stability index for lower EM, a relaxation solution target stability index, a predicted solution target stability index for the first to third prediction units 31 to 33, and a target stability for lower EM for the lower energy operation device 10. Get index and respectively. In addition, the solution quality control unit 36 outputs the acquired relaxed solution target stability index to the planning unit 37 (FIG. 6) and corresponds to the acquired predicted solution target stability index for the first to third prediction units 31 to 33, respectively. It outputs to the 1st-3rd prediction part 31-33, and outputs the target stability index for low-order EM to the low-order energy management apparatus 10 via the communication apparatus 15 (FIG. 5) (S6).

Further, the third predicting unit 33, the demand data accumulated in the demand data accumulating unit 40 (FIG. 6) of the reference data accumulating unit 30 (FIG. 6), the meteorological data accumulated in the meteorological data accumulating unit 41 (FIG. 6), Measured values related to each demand factor accumulated in the demand factor data accumulating unit 42 (FIG. 6), necessary data of the market data accumulated in the market data accumulating unit 43 (FIG. 6), and the second predicting unit. Based on the prediction result of the future energy demand density in the management area of the self-energy operation device given from 32, the future energy demand in the management area is predicted, and the prediction result is output to the planning unit 37 (S7). ..

Further, the planning unit 37 is based on the prediction result of the future energy demand density in the management area of the own energy operation device given from the third prediction unit 33, and the mitigation solution target stability index given from the solution quality control unit 36. Then, a plan for future energy supply in the management area is drafted, and the corresponding external device is controlled according to the drafted plan (S8).

With the above, a series of energy supply planning and control processing is completed. The processes of steps S2 and S3 and the processes of steps S4 and S5 may be executed in parallel, and the process of step S7 may be executed in parallel with the processes of steps S4 to S6. You can

(1-3-2) First Prediction Process FIG. 12 is a specific process content of the first prediction process executed by the first prediction unit 31 in step S2 of the energy supply planning and control process described above with reference to FIG. 11. Indicates.

When the energy supply planning and control process proceeds to step S1, the first predicting unit 31 starts the first predicting process shown in FIG. 12, and first, the reference point adjustment processing unit 50 (FIG. 6) then determines the solution quality. A reference point for predicting future weather in the management area of the self-energy operating apparatus 10 is determined with quality according to the value of the predicted solution target stability index given from the control unit 36 (S10). This determination is performed by determining, as a reference point, a weather grid according to the value of the predicted solution target stability index given by the solution quality control unit 36, which is defined in the first table 56 (FIG. 7).

Subsequently, the data assimilation processing unit 51 (FIG. 6) of the first data assimilation unit 52 (FIG. 6) is executed by the meteorological data and the physical simulation unit that emphasize the reference points determined by the reference point adjustment processing unit 50. A data assimilation process is performed to assimilate the simulation result data of the physical simulation (S11).

Next, the physics simulation unit 55 (FIG. 6) executes the physics simulation of the weather at each reference point based on the weather data of each reference point given from the data assimilation processing unit 51 (S12). Further, the physical simulation unit 55 determines whether the simulation result of the physical simulation converges within a preset threshold value (S13).

Then, if a negative result is obtained in the determination in step S13, thereafter, the processes in steps S11 to S13 are repeated until the simulation result of the physical simulation converges within a preset threshold value.

If a positive result is obtained in the determination in step S13, the mobile observation reference point adjustment processing unit 53 (FIG. 6) selects one of the weather grids from which the weather data was obtained by mobile observation. At this time, a reference point for predicting future weather in the management area of the self-energy management apparatus 10 is determined with quality according to the value of the predicted solution target stability index given by the solution quality control unit 36 (S14).

Next, the sequential data assimilation processing unit 54 (FIG. 6) shows the weather data that emphasizes the reference points determined by the mobile observation reference point adjustment processing unit 53, and the simulation results of the physical simulation executed by the physical simulation unit 55. A data assimilation process for assimilating the data with the above data is executed (S15).

Then, the physics simulation unit 55 performs a physical simulation of the weather at each of the reference points based on the meteorological data of each of the reference points provided by the sequential data assimilation processing unit 54 and the simulation model generated in the physical simulation of step S12. Is executed (S16). As a result, the simulation model generated in step S12 is corrected based on the current meteorological data obtained by the mobile observation. Further, the physical simulation unit 55 determines whether or not the simulation result of this physical simulation converges within a preset threshold value (S17).

Then, when a negative result is obtained in the determination of step S17, thereafter, the processes of steps S15 to S17 are repeated until the simulation result of the physical simulation converges within a preset threshold value. After that, if a positive result is obtained in the determination in step S17, the physical simulation unit 55 outputs the simulation result of the physical simulation to the condition evaluation unit 34, and then the first prediction process ends.

(1-3-3) Second Prediction Process FIG. 13 is a specific process content of the second prediction process executed by the second prediction unit 32 in step S3 of the energy supply planning and control process described above with reference to FIG. 11. Indicates.

When the energy supply plan and control process proceeds to step S3, the second prediction unit 32 starts the second prediction process shown in FIG. 13, and first, the reference point adjustment processing unit 60 (see FIG. 6) of the second data assimilation unit 62. ) Determines the reference point for predicting the future energy demand density of the management area of the self-energy management apparatus 10 with the quality according to the value of the predicted solution target stability index given from the solution quality control part 36 at that time. (S20).

Subsequently, the data assimilation processing unit 61 of the second data assimilation unit 62 attaches importance to the reference points determined by the reference point adjustment processing unit 60, the observation data of each demand factor, and the physical simulation executed by the physical simulation unit 63. A data assimilation process is performed to assimilate the data of the simulation result of (S21).

Next, the physics simulation unit 63 executes a physics simulation of the energy demand density at each reference point based on the observation data of the demand factor at each reference point given from the data assimilation processing unit 61 (S22). Further, the physical simulation unit 63 determines whether or not the simulation result of the physical simulation converges within a preset threshold value (S23).

Then, if a negative result is obtained in the determination of step S23, the processes of steps S21 to S23 are repeated until the simulation result of the physical simulation converges within a preset threshold value. Then, after this, when a positive result is obtained in the determination of step S23, the physical simulation unit 63 outputs the physical simulation result to the regression prediction processing unit 72 (FIG. 6) of the third prediction unit 33 and the condition evaluation unit 34 (FIG. 6). ) And then the second prediction process ends.

(1-3-4) Condition Evaluation Processing FIG. 14 shows specific processing contents of the condition evaluation processing executed by the condition evaluation unit 34 in step S4 of the energy supply planning and control processing described above with reference to FIG.

When the energy supply planning and control process proceeds to step S4, the condition evaluation unit 34 starts the condition evaluation process shown in FIG. 14, and first, the temporal data distribution shape evaluation unit 80 (FIG. 6) performs the first prediction. The magnitude of the temporal change in the predicted value of the meteorological data of each reference point obtained as the simulation result of the physical simulation executed by the physical simulation unit 55 of the unit 31 is, for example, ) And weighted average (S30).

Subsequently, the spatial data distribution shape evaluation unit 81 (FIG. 6) predicts each reference point obtained by the physical simulation executed in each physical simulation unit 55, 63 of the first and second prediction units 31, 32. Based on the value, change in the predicted value between reference points on the same plane parallel to the horizontal direction or between reference points on the same plane parallel to the vertical direction, the variance of the predicted values at the same time section Evaluation is performed by (dispersion value) (S31).

The condition evaluation may be performed according to the degree of multimodality of the distribution of the predicted values instead of the distribution of the predicted values (dispersed value). It may be determined from the number of objects extracted from the spatial data (for example, the number of divisions of the same atmospheric pressure plane of the weather), or a time-series spectrum analysis at the time of prediction of a time section provided every predetermined time (for example, 1 hour). You may evaluate by the power spectrum density (fluctuation intensity) in. In these distribution shape evaluations, calculation processing time increases, but more appropriate evaluation may be performed.

Next, the tag data writing unit 82 (FIG. 6), based on the evaluation result (dispersion value) of the temporal data distribution shape evaluation unit 80 and the evaluation result (dispersion value) of the spatial data distribution shape evaluation unit 81, Future weather conditions obtained by the simulation executed by the physical simulation unit 55 of the first prediction unit 31 and future energy demands obtained by the simulation executed by the physical simulation unit 63 of the second prediction unit 32. The stability with the density condition is determined, and each tag value of the weather condition and the energy demand density condition according to the determination result is written in a predetermined storage area of the external storage device 13 (S32). This completes the condition evaluation process.

In the spatial data distribution shape evaluation, the distribution of the data at the latest time and the distribution of the observation data stored in the reference data storage unit (sample distribution) are evaluated rather than the data distribution of the future time section. Is included. Further, the spatial data distribution shape evaluation may be performed for each time section, and the tag data may be added to each evaluation. In the present embodiment, tag data is added to the evaluation result of the main prediction target time section (for example, 14:00 on the next day).

(1-3-5) Configuration Input Process FIG. 15 shows a specific example of the configuration input process executed by the configuration input unit 35 (FIG. 6) in step S5 of the energy supply planning and control process described above with reference to FIG. The specific processing contents are shown.

When the energy supply plan and control processing proceeds to step S5, the configuration input processing shown in FIG. 15 is started in the configuration input unit 35, and first, via the higher-level energy management apparatus 10 or the input/output interface 14 to be linked. The target stability index for the lower EM set by the user is fetched (S40). Then, the configuration input unit 35 thereafter ends this configuration input process.

(1-3-6) Solution Quality Control Processing FIG. 16 shows specific processing content of the solution quality control processing executed by the solution quality control unit 36 in step S6 of the energy supply planning and control processing described above with reference to FIG. Indicates.

When the energy supply plan and control process proceeds to step S6, the solution quality control unit 36 starts the solution quality control process shown in FIG. 16, and first, the relaxation solution target stability index calculation unit 91 (FIG. 6) causes the condition evaluation. In step S32 of the process (FIG. 14), the tag data writing unit 82 reads each tag value of the weather condition and the energy demand density condition written in a predetermined storage area of the external storage device 13. Further, the relaxation solution target stability index calculation unit 91 reads a corresponding relaxation solution target stability index from the fourth table 90 (FIG. 9) based on the read tag values, and plans the read relaxation solution target stability index. The data is output to the constraint relaxation processing unit 100 (FIG. 6) of the unit 37 (FIG. 6) (S50).

Subsequently, the predicted solution target stability index calculation unit 92 (FIG. 6) reads each tag value of the weather condition and the energy demand density condition in the same manner as the relaxation solution target stability index calculation unit 91, and uses these read tag values as the read tag values. Based on this, the predicted solution target stability indices for the corresponding first to third prediction units 31 to 33 are read from the fourth table 90, and the read predicted solution target stability indices are respectively corresponding to the first to third. The data is output to the prediction units 31 to 33 (S51).

Next, the target stability index calculation unit 93 for the lower EM reads each tag value of the weather condition and the energy demand density condition in the same manner as the relaxation solution target stability index calculation unit 91, and responds based on these read tag values. The target stability index for lower EM is read from the fourth table 90, and the read target stability index for lower EM is transmitted to the associated lower energy management apparatus 10 (S52). This completes the solution quality control process.

(1-3-7) Third Prediction Process FIG. 17 is a specific process content of the third prediction process executed by the third prediction unit 33 in step S7 of the energy supply planning and control process described above with reference to FIG. 11. Indicates.

When the energy supply planning and control process proceeds to step S7, the third predicting unit 33 starts the third predicting process shown in FIG. 17, and first, the reference point adjustment processing unit 70 causes the solution quality control unit 36 to operate. In order to predict future energy demand (electric power demand) in the management area of the self-energy management apparatus 10 with the quality according to the value of the given predicted solution target stability index, the regression model generation unit 71 (FIG. 6) The type of regression model to be generated and the conditions of the regression model are read from the second table 73 (FIG. 8), and the type of the regression model and the conditions of the regression model that are read are output to the regression model generation unit 71 (S60). ..

Subsequently, the regression model generation unit 71 matches the energy demand and the explanatory variable in the management area of the self-energy management apparatus 10 that match the type of the regression model given from the reference point adjustment processing unit 70 and the conditions of the regression model. A regression model between them is generated (S61), and the generated regression model is output to the regression prediction processing unit 72 (FIG. 6).

Next, the regression prediction processing unit 72 acquires the value of the necessary explanatory variable from the simulation result given from the reference data storage unit 30 and the physical simulation unit 63 of the second prediction unit 32, and the acquired value is used by the regression model generation unit 71. The predicted value of the energy demand in the time section (designated time) in the management area of the own energy operation device 10 is calculated by applying the regression model given by Then, the regression prediction processing unit 72 outputs the calculated predicted value of energy demand to the planning processing unit 101 (FIG. 6) of the planning unit 37 (FIG. 6) (S62). With the above, the third prediction process ends.

(1-3-8) Planning Process FIG. 18 shows specific processing contents of the planning process executed by the planning unit 37 in step S8 of the energy supply planning and control process described above with reference to FIG.

When the energy supply planning and control processing proceeds to step S8, the planning processing shown in FIG. 18 is started in the planning unit 37, and first, the constraint relaxation processing unit 100 (FIG. 6) causes the relaxation solution target of the solution quality control unit 36. Based on the value of the relaxation solution target stability index given by the stability index calculation unit 91, the allowable gap ratio from the exact solution for the plan drafted by the plan processing unit 101 is set in the plan processing unit 101 (S70).

Subsequently, the plan processing unit 101, based on the predicted value of the energy demand given from the regression prediction processing unit 72 of the third prediction unit 33, within the range of the allowable gap ratio set by the constraint relaxation processing unit 100, the own energy. An energy supply plan is prepared according to the role of the operation device 10, and the external device is controlled according to the prepared plan (S71). This completes the planning process.

The processing of the plan processing unit 101 is a branch and bound method known as a solution of a constrained optimization problem of a start-stop problem such as when to start and operate power generation (including purchase from the market and negative power), Performs interior point method processing. The ratio from the upper bound/lower bound of the gap (difference between the upper bound and the lower bound), which is the censoring condition for the branch-and-bound method or the interior point method, is used as the allowable gap rate.

In addition to the present embodiment, the conditions are met based on the power generation pattern in which the conditions of demand and weather are classified and the data of what kind of power generation operates to obtain supply under the conditions is classified. You may make it perform the power generation pattern type planning process which makes a power generation pattern a plan result. The number of cases of demand and weather or the number of generation of similar power generation patterns may be increased or decreased.

In this case, when the target stability index from the solution quality control unit is small (a precise plan is intended even though the quality of stability of the solution is low), the constraint relaxation processing unit determines whether the number of cases of demand and weather or Increase the number of generation patterns and increase the target stability index (the stability of the solution is high in quality and the solution of the plan does not change significantly even if there is some calculation error in the demand and weather conditions) The processing unit reduces the number of cases of demand and weather or the number of generation of power generation patterns.

If the number of cases and the number of power generation patterns are large, a power generation pattern closer to the exact solution of the optimization problem can be obtained as the planning result, but the operation of power generation becomes difficult depending on the error of the demand and weather conditions. On the other hand, if the number of cases and the number of power generation patterns are small, an average power generation pattern with a margin is output as the planning result, and the degree of agreement with the exact solution of the optimization problem is low, but demand and weather conditions Power generation can be operated even if there is a slight fluctuation (fringe).

(1-4) Effects of this Embodiment As described above, in the energy management system 1 of this embodiment, the energy management apparatus 10 has the weather condition and the energy demand density condition in the management area, and the higher-level energy management apparatus. Based on the value of the target stability index (target stability index for lower EM) given from 10, the accuracy of energy demand and energy supply planning within the management area is controlled.

In this case, depending on the form of the energy management system 1, the information that can be used in the prediction of future energy demand may be limited to the data measured at the power station of the power system or the primary substation. In this way, forecasting based on a limited physical model that cannot fully utilize physical models of air-conditioning use, traffic use, and industry in the customer's activity area causes a systematic error bias and reduces the prediction accuracy of energy demand. .. In this regard, in the energy management apparatus 10 of the present embodiment, the energy demand is predicted by using the observed values (observation data) of various demand factors such as the road traffic volume in the management area and the communication generation amount. It is possible to prevent the accuracy of forecasting energy demand from decreasing.

If the information that can be used to predict energy demand is limited to the observed values (observed data) recorded in the database, it is not possible to generate a regression model with sufficient reproducibility, and the regression model becomes unstable. I have a problem to do. In forecasting using such an unstable regression model, for example, the regression model may change due to the addition of a few rare observation data such as the impression data on the day of the highest temperature for 10 years. Error variances occur and the accuracy of energy demand forecasting decreases. With respect to this point, in the energy management system 10 of the present embodiment, the first prediction unit 31 also uses the current weather data, so that it is possible to suppress the occurrence of such a problem.

The energy management apparatus 10 of the present embodiment is suitable for various weather and energy demand density conditions such as high bias/low variance prediction and low bias/high variance prediction by performing the above-described processing. It is possible to switch the prediction process such as performing the prediction. Alternatively, it is possible to switch the prediction process to obtain a highly stable solution or a highly rigorous solution, depending on the prediction value that is the solution that the future state is sought using the prediction model (regression model). Become.

Here, the “solution with high strictness” refers to a solution obtained from an accurate model (a model with good reproducibility). For example, in a prediction process under an ideal environment in which an accurate error-free regression model using a large number of explanatory variables is generated, a highly accurate predicted value with a small error (that is, a highly accurate predicted value) Becomes However, it is not always possible to generate a precise model that does not include errors. For example, the regression model adds multiple samples near the boundary of a set of sample data (for example, sample data of the relationship between complicated meteorological data including temperature when the temperature is the highest temperature for 10 years and demand data). If it is done, it will change greatly in proportion to the addition of the sample.

Further, the “solution with high stability” refers to a solution with a highly accurate regression model that does not significantly change with respect to a mathematical condition change such as addition of sample data. The “safe solution” includes some errors, but does not cause a large error as an average when the prediction process is repeated while adding samples each time, and the deviation between each solution is Small (that is, stable). For example, in a prediction process in which a solution of a weighted average model of demand data for the past 10 years is used as a prediction value, even if data for several days is added, the change in the solution becomes small, and the deviation between the solutions is small.

As described above, in the energy management system 1 of the present embodiment, each energy management apparatus 10 determines the accuracy of the energy demand and energy supply plan in the management area based on the weather condition and the energy demand density condition in the management area. In order to control the stability and rigor of the solution (accuracy of the solution and fineness of the time granularity of the solution), which is the quality of the solution that is appropriate for the weather condition and the energy demand density condition in the management area. It is possible to obtain a satisfying solution and to operate energy based on such a solution, so that stable energy supply and adjustment control can be performed.

(2) Second Embodiment FIG. 19 in which parts corresponding to those in FIG. 6 are assigned the same reference numerals is replaced with the energy operation device 10 of the first embodiment, and the energy operation system 1 shown in FIG. The energy management system 110 by 2nd Embodiment applied is shown. The first point of the energy management apparatus 110 is that the second prediction unit 111 predicts the future power generation density in the management area of the self-energy management apparatus 110 instead of the future energy demand density in the management area. The energy management system 10 is largely different from the energy management system 10 of the embodiment, and is otherwise substantially the same as the energy management system 10 of the first embodiment.

In practice, in the energy management apparatus 110, the reference data storage unit 112 is replaced by the market data storage unit 43 (FIG. 6), such as a solar power generation device, a wind power generation device, or a thermal power generation device in the energy operation system 1. Power generation state data for accumulating data (hereinafter, referred to as power generation state data) representing the power generation state of these power generation devices 113 collected from the energy management device 114 that manages each power generation device 113 (generator 6) that generates power. A storage unit 115 is provided.

Then, the physical simulation unit 116 of the second prediction unit 111 executes a physical simulation of the future power generation density (power generation amount per unit area) of the power generation devices 113 in the management area of the self-energy operation device 110, and the data assimilation processing unit. Reference numeral 61 denotes the power generation state data stored in the power generation state data storage unit 115 and the simulation result data of the physical simulation executed by the physical simulation unit 116 (that is, at each reference point determined by the reference point adjustment processing unit 60). Data assimilation processing is performed to assimilate data with the power generation density prediction result).

The prediction of the power generation density is set, for example, in a plane of a mesh of 20 [km]×20 [km] (hereinafter referred to as a mesh plane), which is divided by a weather grid of a weather simulation, for example. This is performed by estimating the power generation density (kw/mesh surface) representing the power generation potential of the power generation device 113 at each time. However, instead of the power generation density, the second prediction unit 111 may predict the future power generation amount within the management area of the self-energy operating apparatus 110.

The simulation result of the physical simulation by the physical simulation unit 116 of the second prediction unit 111 is output to the third prediction unit 117.

The third predicting unit 117, the demand data accumulated in the demand data accumulating unit 40, the meteorological data accumulated in the meteorological data accumulating unit 41, the observation data of each demand factor accumulated in the demand factor data accumulating unit 42, and Based on the required data among the power generation state data accumulated in the power generation state data storage unit 115 and the power generation density in the management area predicted by the second prediction unit 111, the solution quality control unit 36 specifies It is a functional unit having a function of predicting the amount of power generation in the management area of the self-energy management apparatus 10 by quality (stability of prediction according to the value of the predicted solution target stability index given from the solution quality control unit 36), The reference point adjustment processing unit 118, the regression model generation unit 119, and the regression prediction processing unit 120 are provided. In the reference point adjustment processing unit 118, the regression model generation unit 119, and the regression prediction processing unit 120, the CPU 11 (FIG. 5) executes the corresponding program (not shown) stored in the external storage device 13 (FIG. 5). It is a functional unit that is embodied by the above.

Then, the reference point adjustment processing unit 118 determines a reference point in the same manner as the reference point adjustment processing unit 70 (FIG. 6) of the third prediction unit 33 (FIG. 6) in the first embodiment, and determines the determination result. Output to the regression model generation unit 119.

Further, the regression model generation unit 119 holds the same table as the third table 73 (FIG. 8), and in the same way as the regression model generation unit 71 (FIG. 6) in the first embodiment, the self-energy operation is performed. A regression model between power generation and an explanatory variable in the management area of the device 10 is generated, and the generated regression model is output to the regression prediction processing unit 120.

Further, the regression model generation unit 71 acquires the value of the necessary explanatory variable from the simulation result provided from the meteorological data storage unit 41, the demand factor data storage unit 42, and the physical simulation unit 116 of the second prediction unit 11, and the acquired value. Is applied to the regression model given by the regression model generation unit 119 to calculate the predicted value of the power generation amount at the time section (designated time) in the management area of the self-energy operating apparatus 110. Then, the regression prediction processing unit 120 outputs the calculated predicted value of the power generation amount to the planning unit 121.

The planning unit 121 is provided with a control unit 122 instead of the planning processing unit 101 (FIG. 6). The control unit 122 performs feedback control or feedforward control from the power generation device 113. Specifically, the control unit 122 receives a feedback input of the power generation value from the power generation device 113, and sets the power conditioner when the power generation device 113 is a solar power generation device so that a desired amount of power generation is performed. Change the value or the number of solar panels connected to the power conditioner (for example, disconnect the solar panels when the power generation is excessive). Further, when the power generation device 113 is a thermal power generation device, the control unit 122 adjusts the opening degrees of the steam governor and the fuel governor.

The control logic of the control of the power generation device 113 by the control unit 122 uses optimum control, PID (Proportional-Integral-Differential) control, model reference control, fuzzy control, or control by an algorithm known as a neural network. You can

As a process of relaxing the constraint, a process of giving an allowable amount of a control error (that is, a time series of a deviation between a target power generation amount and an actual power generation amount) is performed. In the optimum control logic, parameter adjustment is performed so that allowable optimum control that satisfies the upper limit of the deviation is performed. In the PID control, when the control based on the plan based on the exact solution (the control aiming to make the deviation extremely small) is requested, the process of increasing the parameter of the D coefficient of the PID control is performed. As a result, the control quickly changes the power generation so as to eliminate the deviation at all times. When control based on a plan with a solution for which stability is required is required, control is performed to decrease the parameter of the D coefficient and increase the parameter of the I coefficient, whereby the control result includes a deviation, The change in power generation becomes gentle, and the control stability is improved.

According to the energy management apparatus 110 of the present embodiment as described above, the same effects as those obtained by the energy management apparatus 10 (FIG. 6) of the first embodiment can be obtained.

(3) Third Embodiment (3-1) Configuration of Energy Operation Device of Present Embodiment FIG. 20 in which parts corresponding to those in FIG. 6 are denoted by the same reference numerals is the energy of the first embodiment. The logical structure of the energy management apparatus 130 by 3rd Embodiment applied to the energy management system 1 shown in FIG. 4 instead of the management apparatus 10 is shown. The hardware configuration of the energy management apparatus 130 is the same as that of the energy management apparatus 10 according to the first embodiment described above with reference to FIG. 5 except for a part of the program configuration, and thus the description thereof is omitted here. ..

In the energy management system 130, the processing content of the condition evaluation unit 131 is significantly different from the processing content of the condition evaluation unit 34 (FIG. 6) of the first embodiment. Actually, the condition evaluation unit 131 according to the present embodiment is configured to include the spatial data distribution shape evaluation unit 132 and the tag data writing unit 133.

Then, the spatial data distribution shape evaluation unit 132 of the present embodiment uses the meteorological field (the initial value of the atmospheric state at the designated time section) and/or the predicted meteorological field (the predicted atmospheric state at a certain time). Is classified into several patterns, and a regression model (hereinafter referred to as an error regression model) for calculating the prediction error of the classified weather field and/or predicted weather field pattern is generated for each pattern. It has a function to do. In addition, the spatial data distribution shape evaluation unit 132 determines which of the above-described patterns the current weather field and/or the predicted weather field is classified into, and based on the determination result, the corresponding error regression model is calculated. Regression estimates the error using. Then, the tag data writing unit 133 uses the error thus obtained and the identifier of the pattern of the current weather field and/or the predicted weather field (hereinafter, referred to as a pattern ID) as a tag value, and the external storage device 13 Write to a specified storage area in.

Further, in the energy management system 130, the processing contents of the solution quality control unit 134, the first prediction unit 135, and the third prediction unit 136 are also changed according to the configuration of the condition evaluation unit 131 as described in the first embodiment. Of the solution quality control unit 36 (FIG. 6), the first prediction unit 31 (FIG. 6), and the third prediction unit 33 (FIG. 6).

Hereinafter, the energy management apparatus 130 of this embodiment will be described. In the following, first, specific processing contents of the condition evaluation unit of the present embodiment will be described.

(3-2) Processing Contents of Condition Evaluating Unit According to this Embodiment FIG. 21 shows specific processing contents of a series of processes executed by the condition evaluating unit 131 according to this embodiment described above. In the present embodiment, the prediction data of each meteorological element acquired from the meteorological prediction system includes a numerical fluid analysis (CFD) of the atmosphere, a thermodynamic analysis of heat exchange between the atmosphere and the surface and space, and between the atmosphere and the surface of the earth. It is calculated by friction mechanics analysis, thermo-fluid analysis based on thermal analysis of solar radiation and earth radiation, and meteorological numerical prediction that includes data assimilation of analysis values of meteorological observation data.

Specifically, data assimilation is performed to correct the analysis value so as to reduce the error between the observed value and the analysis value (first estimated value) of the meteorological numerical prediction performed in the previous prediction cycle, and the prediction calculation is performed. The initial value of the atmospheric state, which is the analysis value (second estimated value obtained by the analysis increment) of the meteorological data related to the start time of, is calculated, and the numerical fluid analysis and the thermodynamic analysis are performed with respect to the initial value of the atmospheric state. It is calculated by performing time integration (time evolution) using the calculus formula of (1) and simulating the state value at an arbitrary time.

Specific processing contents of the condition evaluation unit 131 in each step will be described. First, the spatial data distribution shape evaluation unit 132 of the condition evaluation unit 131 acquires the meteorological data (initial value and/or predicted value of meteorological numerical value prediction) accumulated in the meteorological data accumulation unit 41 over the past ( S80).

Then, the spatial data distribution shape evaluation unit 132 classifies each of the weather field pattern of the initial value of the weather numerical prediction and the predicted weather field pattern into some patterns based on the acquired weather data. (S81, S82). However, step S82 can be omitted when the condition evaluation process is performed only on the weather field, and step S81 can be omitted when the condition evaluation process is performed only on the predicted weather field. .. The predicted time evaluated by the spatial data distribution shape evaluation unit and the predicted time finally desired to be evaluated need not be limited to the same time, and may be different times.

Next, the spatial data distribution shape evaluation unit 132 causes the weather field pattern of the initial value of the weather numerical value prediction and/or the predicted weather field pattern and the error of the weather numerical value prediction that has occurred in the past (error of temperature, insolation). Generates a regression model (hereinafter referred to as error regression model) to detect the relationship with the magnitude of the value of the elements included in the numerical weather forecast such as the error of flux, the error of humidity, and the error of wind direction To do. The relationship between the pattern of the meteorological field and the magnitude of the error of the numerical weather prediction is generated and detected by the regression model, the neural network model, the SVM (support vector machine) model, and the decision tree model such as CART. You may do it.

Specifically, for example, as shown in FIG. 27, the prediction error amount for each pattern is aggregated, and the aggregated prediction error amount for each pattern is shown in FIG. Such a sample variance graph is generated as an error regression model. When there are a plurality of error magnitudes that have occurred in the past for one pattern, the maximum error and the average value of the error are used to detect the relationship with the prediction error amount (S83). If the adjustment process of step S97 described later allows the adjustment process in which the magnitude of the error generated for each pattern is uniform to sufficiently converge, an error regression model with the average error is created to perform good error prediction. You may do it.

After that, the spatial data distribution shape evaluation unit 132 performs pattern matching with the current meteorological data being distributed, and the meteorological field in which the current meteorological data (initial value and/or predicted value) is classified or the predicted weather. Acquire the pattern ID of the field pattern. Specifically, a process of mapping the current meteorological data to a latent space of a self-organizing map (hereinafter, referred to as SOM latent space) described below and acquiring a pattern ID of a pattern grouped in the SOM latent space is performed. Perform (S84).

Subsequently, the spatial data distribution shape evaluation unit 132 estimates the error in the weather numerical value prediction using the error regression model corresponding to the pattern ID acquired in step S84 (S85).

After that, the tag data writing unit 133 of the condition evaluation unit 131 determines at least one of the data of the estimation error obtained by the process of step S85 and the pattern ID of the pattern of the classification destination of the current weather data. For example, the data is written to a predetermined storage area in the external storage device 13. Alternatively, if the estimation error is larger than the average value (or if it is larger than the value obtained by adding the value of the variance σ of the error amount to the average value of the error amount), write the degree of inferiority as the value of the inferiority tag of the analysis. Yes (S86).

In step S82, an error regression model showing a relationship with the prediction error of the third prediction unit (demand) to be finally known is generated instead of the error in the weather numerical prediction, and in step S85, the third prediction is performed. It is also possible to configure so as to estimate the worst error of the demand forecast value output by the section.

FIG. 22 shows the processing content of more specific processing (hereinafter, referred to as pattern classification processing) of the spatial data distribution shape evaluation unit 132 in steps S81 and S82. Here, it is assumed that the pattern classification of the weather field is mainly performed from the viewpoint of the potential temperature.

The "warm level" is a scale indicating the substantial warmth of the air at different altitudes. According to the theory of thermodynamics, the air mass in the upper layer has a low atmospheric pressure in the surroundings, so when it is adiabatically moved to the lower layer (atmospheric pressure is around 1000 hPa), the temperature rises under the influence of atmospheric pressure. Thus, if there is an air mass of the same temperature in both the upper layer and the lower layer, it should be considered that there is substantially a warm air mass in the upper layer. In addition, the humidity is high when air contains water vapor as a component (water (chemical formula H2O) has a phase of liquid and gas, and vaporized water is called water vapor). When the water vapor changes its phase to a liquid, it gives off heat, and conversely liquid water becomes a gas and requires external heat energy. Therefore, if the air mass containing water vapor and the air mass not containing water vapor (however, liquefied water may float as water droplets) have the same temperature, the air mass containing water vapor is the better. Should be considered substantially warmer. In meteorology, as a measure of such a temperature level, a theoretical level of a temperature level for dry air (often called simply a temperature level) or a theoretical temperature level for wet air is used.

In the present embodiment, the temperature and humidity data of the air mass at each altitude are used as the data related to the temperature level. Note that the present invention is not limited to this embodiment, and the theoretical value of the temperature of the dry gas or the equivalent temperature may be used. Further, as the vertical arrangement direction data, not only the temperature level but also the items of meteorological data such as the value of the wind direction, the value of the ascending and descending airflow, the value of the geopotential altitude, and the like may be created.

Specific processing contents of each step will be described. When the spatial data distribution shape evaluation unit 132 proceeds to step S81 or step S82 of FIG. 21, this pattern classification processing is started, and first, for a predetermined period (for example, for each of the four seasons, for a year, for a certain period of daytime, for a certain period of time). From the meteorological data of each time section accumulated at night in a predetermined time zone (for example, the time zone when the global weather forecast is updated, the meteorological data distribution time zone at 6:00 am), the point for each of the horizontal grids From the data items for each altitude included in the data, the temperature-related data described below is extracted to generate a vertical arrangement data set of meteorological data (S90).

Subsequently, the spatial data distribution shape evaluation unit 132 classifies the generated set of vertical arrangement data of weather data for each point into some patterns (S91 to S97).

Specifically, the spatial data distribution shape evaluation unit 132 first performs processing of a data set to a node of the SOM latent space by a process known as a self-organizing map (hereinafter referred to as a self-organizing map process). Each data is mapped (S91).

Then, the condition evaluation unit connects the nodes of the SOM latent space by using a predetermined threshold by self-organizing map processing, sets an appropriate separation threshold, and performs pattern classification on the SOM latent space, and then, By reverse mapping the data from the SOM latent space, each data of the vertical arrangement data set of the meteorological data prepared in S90 is pattern classified. Further, the spatial data distribution shape evaluation unit 132 gives a tag (hereinafter, referred to as stability ID) to each pattern of the classified vertical arrangement data of the air mass (S92).

Further, the spatial data distribution shape evaluation unit 132 maps the map data representing the stratified state of the atmosphere in which (corresponding to) the corresponding stability IDs are associated (displayed) with the respective grids at the horizontal points on the respective time planes ( Hereinafter, this will be referred to as stratified state data) is generated corresponding to each of the time section data of the distributed and accumulated weather (S93).

In the present embodiment, the stability ID is given according to the value of the atmospheric stability with respect to the pattern. "Atmospheric stability against a pattern" means the average value (or maximum value) of the vertical stability index of each atmosphere (for example, the Schwarter stability index of meteorology) of the vertical arrangement data of air masses with the same pattern. Value). As another method of assigning the stability ID, an air mass corresponding to a reference vector (indicating a virtual representative value of a pattern) from a node of the SOM latent space to a data space (also referred to as an observation space) of data to be classified The atmospheric vertical stability for vertical placement may be used.

By adding an ordinal number as the stability ID to the pattern obtained by extracting the shape of the arrangement of air lumps in the vertical direction in this manner, the stability IDs of the arrangement patterns having similar shapes have close values. For this reason, when the stratified state of the atmosphere on the horizontal plane is converted into data in the shape of the arrangement of air masses in the vertical direction for the atmosphere at each lattice position on the horizontal plane, a minute difference in the accounting results in a large change in stability ID. There is no practical problem, and there is the practical effect that good horizontal data can be created.

Subsequently, the spatial data distribution shape evaluation unit 132 maps the stratified state data generated in step S93 to the second SOM latent space by the self-organizing map process similarly to the process of S92 (S94), The stratified state data of the atmosphere is pattern-classified in the SOM latent space of 2 (S95).

Then, the spatial data distribution shape evaluation unit 132 determines whether or not each pattern has sufficient resolution (S96), and when a negative result is obtained, the pattern is set so that each pattern has sufficient resolution. The value of the above-mentioned separation threshold used in step S92 is adjusted so as to increase or decrease the number of classifications (S97).

Then, the spatial data distribution shape evaluation unit 132 returns to step S92, and then repeats the processing of steps S92 to S97 until a positive result is obtained in step S96. Then, when the spatial data distribution shape evaluation unit 132 eventually obtains a positive result in step S96 by finishing the pattern classification of the weather field or the predicted weather field into some patterns with sufficient resolution, this pattern classification process is performed. When it is finished, the process returns to the process of FIG.

In the pattern classification process of S96 and/or S92, an identifier (pattern ID) is given to the pattern generated from the data. The identifiers are assigned such that the ordinal numbers of the identifiers that are close to each other have similar patterns.

Specifically, the degree of difference between adjacent nodes of the SOM latent space (in one of the map color method algorithms, the observation data of the observation space indicated by the node of the SOM latent space and the observation data of the adjacent node of the SOM latent space indicate The distance from the observation data can give different degrees of adjacent nodes) and the separation threshold R are compared to try to connect the nodes. R is changed from RMAX, which is a different degree value of the maximum adjacent node, to 0 (for example, changed in 5 steps), and trials are performed until the number of nodes that are gradually connected is classified into a small group. A semi-ordered tree is given to each node in the SOM latent space by a semi-ordered tree based on the classification in stages. The pattern ID is given according to the partial order relationship. By assigning the pattern ID in this way, the nodes that are largely classified by the root due to the partial-order tree structure have significantly different properties, and the nodes that are classified by the terminal branches have similar properties.

According to this embodiment, since patterns having similar properties (for example, a weather pattern corresponding to a so-called winter/summer weather arrangement) have close pattern IDs, the prediction processing in the post-processing, And/or the error regression model can be generated with good quality. This is because even if the number of data classified into each pattern is heterogeneous (the number of data is sparse and dense, it is sparse), it can be processed by Gaussian process regression and/or sparse modeling such as histogram method. This is because it is possible to generate a good analytical model.

Note that, instead of the processing of the present embodiment, the value of the vertical stability of the atmosphere may be used as the stability ID tag of the point. In meteorology, as an evaluation of the vertical arrangement data of the temperature level, the arithmetic difference of the values between the data of the upper level temperature and the data of the lower level temperature (equivalent temperature or dry gas temperature) is often used. It is used as an evaluation value as stability. If the stability is high, it is difficult for updrafts from the lower layers to occur due to the laws of thermodynamics.

In addition to the present embodiment, the spatial data distribution evaluation unit 122 may perform temporal data distribution shape evaluation that integrates the evaluation results of the meteorological field at each time section. It is possible to perform temporal data shape evaluation that there is a peculiar phenomenon during development that the degree of inferiority increases with the passage of time.

23A to 23G show examples of the result of pattern classification of the vertical arrangement of the air mass in the data regarding the temperature level. The vertical axis represents the atmospheric pressure altitude, and the horizontal axis represents the calculated equivalent temperature conversion value. Further, FIG. 24 shows an example of the result of pattern classification of the stratified state data. Further, FIG. 25(A) shows an example of the error regression model described above, and FIG. 25(B) exemplifies the relationship between the pattern of the graph stratified state data representing the number of samples of each pattern and the sample variance of the prediction error. An error regression model is generated using this graph. The relationship between the pattern and the error may be detected after normalizing the prediction error amount of each pattern by using the number of samples for each pattern.

(3-3) Flow of Processes in Energy Operation Device of Present Embodiment FIG. 26 is a flow of a series of processes regarding energy supply planning and control executed in the energy operation device 130 of the present embodiment (energy supply plan). And control processing). This energy supply plan and control process are executed regularly.

In the energy management system 130 according to the present embodiment, first, the processing described above with reference to FIGS. 21 to 25 and 27 is executed in the condition evaluation unit 131, and the tag data writing unit 133 of the condition evaluation unit 131 outputs the processing result. At least one of the tag value of the weather condition (estimated amount of error in weather numerical value prediction) and the pattern ID of the current weather field and/or the predicted weather field is written in a predetermined storage area of the external storage device 13 (S100). ).

Then, the numerical analysis parameter updating unit 137 of the solution quality control unit 134 determines the first from the tag value of the weather condition written in the external storage device 13 and the pattern ID of the current weather field and/or the predicted weather field. The parameters used for the reanalysis processing of the meteorological phenomenon performed by the 1 prediction unit 135 are corrected (S101).

In the case of the present embodiment, the numerical analysis parameter updating unit 137 determines the parameters α1, α2 and/or the ratio α1 and α2 indicating the magnitude of the observation error and the background error, which are parameters of data assimilation. , Change according to the weather field. Conventionally, α is set to a fixed standard value. Change α according to the weather field. Specifically, for the time you want to predict (for example, 24 hours later, a time section in 30-minute increments 6 hours before and after the time t at which prediction is performed using meteorological numerical prediction data such as demand and power generation) Process of increasing the value of α according to the size of the “numerical analysis error estimated from the pattern ID” (for example, adding a value obtained by multiplying the value of the numerical analysis error by a predetermined value β to standard α). Apply. Further/or, a process of increasing the value of α is performed according to the magnitude of the numerical analysis error estimated from the pattern ID of the weather field, which is the initial value of the weather analysis.

The numerical analysis parameter updating unit 137 can also update the parameters (these parameters are held as a constant table of the physical simulator PG) of the physical simulation unit 55 for time evolution of atmospheric conditions.

The predicted value of the wind speed data (the average value of the predicted values in the area to be predicted, and/or the predicted value of the point where you want to make a final prediction such as demand) was determined to be an excessive weather condition. In this case, the value of the parameter K indicating the coefficient of friction between the ground surface and the atmosphere is slightly increased (for example, increased by 0.01%). When it is determined that the predicted value is an excessively small weather condition, the value of the parameter K indicating the friction between the ground surface and the atmosphere is slightly reduced (for example, 0.01% reduction).

It is known that if the friction coefficient between the ground surface and the atmosphere is mistakenly made small, the wind speed will diverge or converge to an unrealistic value when simulating a long time evolution. There is. By dynamically changing the parameters of the physical simulation, the error can be suppressed by reanalyzing the numerical analysis of the meteorology even under unique meteorological conditions.

Similarly, when the predicted value of the ground surface temperature and/or the predicted value of the atmospheric temperature is determined to be an excessively large meteorological condition, terrestrial radiation (radiant energy received from the sun by the earth, Parameter M of the total energy emitted from the surface of the earth and the atmosphere to the outer space as infrared rays. Is slightly increased (for example, increased by 0.01%) and/or the parameter of solar radiation (total energy of electromagnetic waves emitted from the sun) is slightly reduced. On the contrary, when the predicted value of the ground surface temperature and/or the predicted value of the atmospheric temperature is determined to be an excessively small meteorological condition, the earth radiation (radiated energy that the earth receives from the sun, The parameter M of the total energy emitted from the surface of the earth or the atmosphere to the outer space as infrared rays) is slightly decreased (for example, increased by 0.01%), and/or the parameter of the solar radiation (the total energy of electromagnetic waves emitted from the sun) is slightly increased.

If the parameters of the Earth's radiation are mistakenly set to be large relative to the magnitude of the solar radiation, the temperature of the ground surface and the atmosphere will diverge or converge to unrealistic values during the simulation of long-term time evolution. It is known to do. If the parameters of the earth's radiation are mistakenly made small, simulations of long-term time evolution will cause the temperatures of the ground surface and the atmosphere to converge to extremely low temperatures, unrealistic values, and periodic changes such as the rotation of the earth. It is known that it vibrates to a periodic predicted value with no information amount that reflects only this. By dynamically changing the parameters of the physical simulation, the error can be suppressed by reanalyzing the numerical analysis of the meteorology even under unique meteorological conditions.

Then, the data assimilation processing unit 139 of the first data assimilation unit 138 of the first prediction unit 135 uses the governing equation of the physical current state of the atmosphere to simulate the time evolution of the state of the atmosphere. Initial data of grid data is generated (S102).

Specifically, the initial data obtained by the weather numerical solution prediction (and/or the initial value data obtained by the weather reanalysis and/or the predicted value which is the analysis result in the previous weather analysis cycle) is first As an estimated value, an analysis value of an initial value is calculated between the acquired observation data (observation value) and the first estimation value is updated (this process is generally called an analysis increment, which may be referred to as a conditional increment).

The update of the first estimated value is performed by inputting the first estimated value which is a previous analysis value and the observation data and performing data assimilation with the adjoint equation of the governing equation relating to the atmospheric state by a process known as a variational method. It is determined by the ratio α value related to the setting of “observation error” and “background error” (error of the first estimated value) given when performing. At this time, if the observation error is set smaller than the background error, the analysis increment result due to data assimilation will be closer to the observation data side. Conversely, if the observation error is set larger, the analysis increment result will be the first estimated value side. Stop by.

Based on the initial values obtained in this way, re-analyzing the meteorological phenomenon (the re-analyzing may subdivide the grid spacing of the analysis rather than the meteorological analysis by an external server) through a simulation of time evolution Get an estimate of the state of.

Subsequently, the spatial data analysis shape evaluation unit 122 of the condition evaluation unit 131 evaluates the predicted condition of the meteorological field similar to step S100 with respect to the prediction result of the time evolution of the atmospheric state related to the weather ( S104), and thereafter, the spatial data analysis shape evaluation unit 122 compares the error amount of the reanalysis result with the error amount before the reanalysis, and performs a convergence determination as to whether or not the error reduction amount is less than a specified value. (S105). In this convergence judgment, if the judgment result that the re-analysis result has not converged sufficiently is obtained, until the re-analysis result converges below the threshold value, or until the specified number of iterations are completed, The process from the solution quality control process (S101) to the condition evaluation process (S104) is repeated.

Through the processing of steps S101 and S102 described above, the differential equation (dominant equation) used in the physical simulation for predicting the time evolution of the atmospheric state due to the peculiar meteorological phenomenon and the actual physical phenomenon are completely Even when the values do not match (that is, when the background error is large), the results of data assimilation (processing of initial value generation) are improved by adapting to the increase in the background error due to a peculiar meteorological phenomenon. This has the effect of improving the accuracy of reanalysis.

Furthermore, the processing of reducing the value of α is performed by using the tag information of the data quality such as the lack of meteorological observation data, the delay of the observation data delivery, or the freshness of the observation data (elapsed time after the observation) ( For example, the standard α value multiplied by 0.9 is used as the α value). As a result, in exceptional circumstances where the observation error is increasing, the result of the analysis increment generated by the initial value is larger than that in the normal case when the numerical prediction result (first An analysis increment is performed to output a value close to (estimated value). As a result, it is possible to avoid the result of data assimilation being extremely deteriorated by the observation data deteriorated due to missing observation data, delay, or deterioration of freshness, and to obtain the data assimilation result of constant quality (analysis increment result). There is an effect that can be obtained.

In addition, the meteorological numerical prediction data distributed from the meteorological organization is an area for analyzing the atmospheric condition of a specific area such as the Japan area by performing a global model analysis for analyzing the atmospheric condition of the entire earth, for example, every 24 hours. For the model analysis, input the analysis result of the global model and the latest observation data (however, depending on the observation item, it may be observation after 1 hour or observation after 6 hours), for example, 3 hours The freshness of the data used in the prediction cycle is different, such as every time. As a result, the characteristics of the error included in the meteorological data (for example, the magnitude of the observation error and the background error) also differ depending on the distribution time (the announcement time of the weather forecast). In the present invention, the standard values of α1 and α2 in the reanalysis may be set for each distribution time. This allows good reanalysis.

Also in the numerical analysis of the second prediction process (S103), similarly, data assimilation using parameters of the magnitude of the observation error and the background error may be performed, and the observation error due to the lack of the observation data increases. Sometimes it is possible to properly estimate the energy demand density.

Subsequently, the configuration input processing described above with respect to step S5 of FIG. 11 is executed by the configuration input unit (S106). In addition, thereafter, the predicted solution target stability index calculation unit 92 (FIG. 6) of the solution quality control unit 134 reads the tag value of the estimated value of the weather prediction analysis error in the weather condition, and the third prediction unit The predicted solution target stability index is output (S107). The predicted solution target stability index may be a real number value of the estimated value of the meteorological analysis error, or may be converted into a positive number value of 1 to 5.

Next, the third prediction unit 136 executes the third prediction process described above with reference to FIG. 17 (S108). In addition, in the third prediction process of the present embodiment, the prediction of the power demand is predicted as F(W)+G(t), which is the sum of the weather-dependent term F(W) and the non-climate-dependent term G(t). To do. F(W) is a formula of the above-mentioned Lasso regression model and Ridge regression model, and G(t) is a fixed value set statistically at each time t.

The reference point adjustment processing unit 70 determines the future energy demand (electric power) in the management area of the own energy management apparatus 10 with the quality according to the value of the predicted solution target stability index given from the solution quality control unit 36 at that time. In order to predict the (demand), the type of regression model to be generated by the regression model generation unit 71 of FIG. 6 and the conditions of the regression model are read from the second table 73 of FIG. The conditions of the regression model are output to the regression model generation unit 71 (S60 in FIG. 17). Especially when generating a regression model, when the tag value of the weather condition inferiority is large, the weather element used as the explanatory variable of the regression model is limited to the temperature, and when the weather condition is not inferior, the regression model is explained. Without limiting the meteorological elements used for variables, you can use items such as humidity, solar radiation, atmospheric suspended matter amount, infrared intensity, ultraviolet intensity, cloud amount, weather (clear, cloudy, rain, snow), and weather at each time section. Meteorological items of the meteorological data of a time section having a large inferiority tag value may be limited from the inferiority tag value of the numerical prediction.

In addition to the present embodiment, the demand is modeled by the weather-dependent model F(W) and the non-climate-dependent model G(t), and the ensemble synthesis (a×F(W)+(1-a)×G( There may be a case where the demand is predicted by the value of (t)) (first estimated value). The value of a may be corrected from the standard value based on the tag value of the degree of inferiority of the weather condition. For example, when the standard a is 0.7 and the degree of inferiority of the weather condition is large, a is reduced to 0.5. Alternatively, a is reduced in proportion to the estimated error value of the numerical weather analysis. The ensemble combination is not limited to linear combination, and may be non-linear combination, or the function Z(F, G) used for non-linear combination may be estimated by the variational method. The regression model generation unit 71 may perform Gaussian process regression to create a stochastic regression model between climate and demand.

By limiting the meteorological data to be used in this way, it is possible to prevent the accuracy of the prediction using the data from being greatly deteriorated when the error of the meteorological numerical prediction data is large.

After that, as described above with reference to FIG. 17, the regression prediction processing unit 72 acquires the values of the necessary explanatory variables from the simulation results provided from the reference data storage unit 30 and the physical simulation unit 63 of the second prediction unit 32, and acquires the values. The predicted value of the energy demand in the time section (designated time) in the management area of the own energy management apparatus 10 is calculated by applying the calculated value to the regression model given by the regression model generation unit 71. Then, the regression prediction processing unit 72 outputs the calculated predicted value of energy demand to the planning processing unit 101 (FIG. 6) of the planning unit 37 (FIG. 6) (S62). With the above, the third prediction process ends.

By doing so, processing using numerical analysis data based on numerical analysis condition estimation that estimates errors related to numerical analysis of weather (for example, using natural energy or outside air that depends on power demand or weather) Output increment processing that corrects the result of (the prediction of the power generation amount of thermal power generation) is achieved.

(3-4) Effects of this Embodiment According to the energy management system 130 of this embodiment as described above, the physical characteristics of the meteorological field of the arrangement of air masses in the atmosphere and the numerical fluid analysis of the weather prediction. It is possible to predict the analysis error while taking into account the characteristics of (1) and perform re-analysis to reduce the predicted analysis error. Therefore, according to the energy management system 130 of the present embodiment, in addition to the effects obtained by the first embodiment, the analysis error of the numerical analysis and/or numerical prediction is estimated, and effective reanalysis is performed. The effect of obtaining can be obtained.

(4) Fourth Embodiment (4-1) Power Generation Reserve In “EM1” (2AA) and “EM2” (2AB) of FIG. 1, it is required in advance to issue a generation command for positive watts and negative watts. Posiwatts and Negawatts are predicted. These positive watts and negative watts are prepared for a shortage of planned power supply against expected demand, and are called down adjustment power, up adjustment power, or simply adjustment power or power reserve. This reserve capacity is caused by an unexpected increase/decrease in demand at the time of initial planning and an unexpected change in weather-dependent power generation such as solar power generation and wind power generation. Therefore, the unexpected demand and the amount of power generation at the time of the initial plan are caused by the meteorological numerical forecast and/or the change of the demand density at the time of the initial plan. It is possible to create a model (for example, a regression model) of the relationship with the estimated amount of error in various numerical analyzes such as weather, demand, and power generation. Using the created relational model between the numerical analysis error and the power reserve, and the above-described estimated amount of the numerical analysis error (output of step S85 in FIG. 21), the required power reserve can be estimated/predicted. .

(4-2) Prediction of price unit price of reserve power generation The market price between the orderer and the contractor is determined by trading on the adjustment power exchange 2BC. Since the contract price has a causal relationship between the demand associated with a meteorological phenomenon and the physical quantity of power generation, the price depends on the error of numerical analysis such as the weather. Adjustable power Exchange rate unit price (unit price per kW or unit price including adjustment duration) and estimated amount of error in various numerical analyzes (related to weather, demand, power generation) A model of the relationship (eg, regression model) can be created. The transaction price can be estimated/predicted by using the created relational model of the numerical analysis error and the transaction price and the above-described estimated amount of the numerical analysis error (output of step S85 in FIG. 21).

(5) Other Embodiments In the above-described first, second and third embodiments, the data of meteorological elements (meteorological data) and the data of living activities (used by each energy management device 10, 110) ( The case where the observation data of each demand factor) is not distinguished has been described, but the present invention is not limited to this, and for example, in the energy management system 1, the upper energy management devices 10 and 110 and the lower energy management devices. The energy demand may be predicted by using the meteorological data of the weather grid of altitudes different from 10, 110 and the observation data of the demand factors at different points. By doing so, it is possible to avoid erroneous predictions of the energy operation devices 10 and 110 at the same time, and it is possible to perform stable energy operation as the entire energy operation system 1.

Further, in the above-described first and second embodiments, in the energy operation system 1, the energy operation devices 10 and 110 managed by the regional power transmission company and having the role of eliminating the deviation between the demand and supply of energy, The case has been described in which the weather data used by the energy management devices 10 and 110 that are managed by the retail power generation company and that adjust the supply capacity are not distinguished, but the present invention is not limited to this, and the former energy management is used. The devices 10 and 110 use the weather data of the weather grid from the upper layers, and the latter energy operation devices 10 and 110 use the weather data from the ground surface to construct an energy operation system to predict energy demand. You can By doing so, the former energy management device 10, 110 can stably predict the weather condition of the entire region, and the latter energy management device 10, 110 can accurately detect the increase or decrease in the demand of large-scale customers. Therefore, it is possible to operate the generator economically.

Further, in the above-described first embodiment, the data assimilation processing unit 51 of the first data assimilation unit 52 of the first prediction unit 31 performs the data assimilation processing that emphasizes the reference points determined by the reference point adjustment processing unit 50. Although the case has been described above, the present invention is not limited to this, and, for example, the data assimilation processing unit 51 determines, based on the prediction result of the second prediction unit 32, a point where the energy demand density in the management area is high. You may make it determine and may perform the above-mentioned data assimilation process about these points. By doing so, the accuracy of forecasting and estimating the values of meteorological elements such as temperature, insolation, humidity, wind speed, wind direction, and atmospheric pressure is improved in areas where energy demand is high, and the forecast of energy demand that depends on meteorological elements is improved. The accuracy can be improved.

Similarly, in the above-described second embodiment, for example, the data assimilation processing unit 51 of the first prediction unit 31 selects a point with high power generation density in the management area based on the prediction result of the second prediction unit 32. You may make it determine and may perform the above-mentioned data assimilation process about these points. By doing so, the accuracy of prediction/estimation of the value of the meteorological element in the area where the energy generation density is high can be improved, and the prediction accuracy of energy generation depending on the meteorological element can be improved.

Furthermore, in the above-described first and second embodiments, the solution quality control unit 36 specifies the quality of the predicted solution to the first to third prediction units 31 to 33 with only one numerical value (predicted solution target stability index). However, the present invention is not limited to this, and, for example, the quality of the predicted solution can be set to a plurality of values such as stability, rigor, accuracy, remedy speed, solution time granularity, and solution spatial granularity. The solution quality control unit 36 specifies the quality of each of these items to the first to third prediction units 31 to 33, and the first to third prediction units 31 to 33 specify the quality of each of these items. The predicted solution may be calculated so as to obtain the predicted solution according to the designated quality.

Further, in the above-described first embodiment, the solution quality of the predicted solutions of the first to third prediction units 31 to 33 is controlled based on the weather condition and the energy demand density condition within the management area of the energy management system 10. In the second embodiment, the solution quality of the predicted solutions of the first to third prediction units 31, 111, 117 is controlled based on the weather condition and the power generation density condition in the management area of the energy management system 110. However, the present invention is not limited to this, and the configurations of the first and second embodiments are combined to create a weather condition, an energy demand density condition, and a power generation density condition in the management area of the energy management system. You may make it control the solution quality of the prediction solution in the 1st-3rd prediction part based on three conditions.

Moreover, in the above-mentioned 1st-3rd embodiment, although the prediction system which outputs a numerical-analysis value and a numerical-analysis error in the form which each energy management device utilizes was described, this invention is not restricted to this. , Other systems that predict numerical analysis values and/or numerical analysis errors.

For example, it can also be used in a system for controlling indoor air conditioning by controlling the opening and closing of ventilation windows at night. The night purge, which opens the ventilation window at night and releases the indoor heat accumulated during the day, may have an adverse effect when the outside air temperature is high or the humidity is high. Therefore, the weather of the outside is re-analyzed, the distribution of the heat fluid inside the room is analyzed, and the effect of the air exchange inside and outside the room is predicted to perform the control. Especially when the prediction error is large, a process of changing the ventilation control execution time to a time period in which the prediction error is small is performed.

Further, in the above-described first to third embodiments, the case where the numerical analysis target is the weather is described, but the present invention is not limited to this, and the calculation by the mechanical formula, the structural analysis, the electromagnetic wave analysis, It can be widely applied to analysis of electric circuits of electric power systems.

In the conventional simulation for prediction by numerical analysis, the characteristics of the dynamic formula used in the simulation that is derived a priori from the theoretical formula are not sufficiently taken into consideration, and the simulated state when the time evolution in the simulation is advanced No method has been provided for classifying when and how convergence, divergence, vibrations, etc. of the value of ω occur from the analysis result and/or the data situation regarding the initial value. The present invention classifies patterns of spatial distribution of initial values of analysis and/or data of output results of analysis and/or patterns of temporal distribution. It provides a method of formulating the relationship between the analysis error that has accumulated as historical data and the error of the predicted value that you want to finally know. Thereby, it is possible to determine whether the initial value and/or the analysis output value is similar to the pattern in which the error has been generated in the past, and to estimate the error generated by the simulation.

INDUSTRIAL APPLICABILITY The present invention provides various energy management systems each composed of a plurality of energy management devices that predict demand and/or supply of energy in a management area, and operate the energy in the management area based on the prediction result. Can be widely applied.

1, 1A to 1D... Energy operation system, 2, 10, 110, 130... Energy operation device, 4... Consumer, 5... Power plant, 6... Generator, 11... CPU, 31,135 ......First prediction unit, 32,111 ......Second prediction unit, 33,136 ......Third prediction unit, 34,131 ......Condition evaluation unit, 35 ......Configuration input unit, 36,134 ......Solution Quality control unit, 37, 117... Planning unit, 56... First table, 73... Second table, 81, 132... Spatial data distribution shape evaluation unit, 82, 133... Tag data writing unit , 90... Third table, 102... Fourth table, 113... Power generation device, 137... Numerical analysis parameter updating unit.

JP, 2016-93016, A JP, 2008-31683, A

Claims (12)

In a numerical analysis device that performs numerical prediction,
Data of initial conditions of numerical analysis, and/or a pattern classification unit that classifies the analysis result data of the numerical analysis into a plurality of patterns,
A detection unit that detects the relationship between each pattern and the inferiority of the prediction result,
And a prediction inferiority estimation unit that estimates the inferiority of the prediction result of the numerical prediction target based on the detection result of the detection unit for the pattern into which the numerical prediction target is classified. Analyzer. The numerical analysis device according to claim 1, wherein
Inferiority data of the result of the first prediction regarding the state of the system targeted for the numerical prediction, and inferiority of the result of the second prediction based on the result of the prediction regarding the state of the system targeted for the numerical prediction. A numerical analysis device, comprising: a predictive inferiority display unit that displays at least one of the above data. The numerical analysis device according to claim 2, wherein
The second data based on at least one of the data of each of the patterns of the initial condition data, the data of each pattern of the analysis result, and the inferiority data of the prediction result of each pattern. Prediction is performed and/or data of each of the patterns of the data of the initial condition, data of each of the patterns of the analysis result, data of inferiority of the prediction result of each of the patterns, and the second prediction A numerical analysis device, comprising: an information processing unit that estimates a reserve amount to be secured in case of a prediction failure, based on at least one of the inferiority data included in. The numerical analysis device according to claim 1, wherein
The prediction target is a meteorological phenomenon,
Data of the initial condition of the numerical analysis for the meteorological phenomenon, and a meteorological data acquisition unit for acquiring at least one of the analysis result data,
The pattern classification unit,
A numerical analysis device, wherein the data of the initial condition and/or the data of the analysis result of the numerical analysis for the meteorological phenomenon acquired by the meteorological data acquisition unit is classified into a plurality of the patterns. The numerical analysis device according to claim 4, wherein
The inferiority of the prediction result is an error in the numerical weather prediction, which is the time evolution of the numerical weather analysis,
The degree of inferiority of the prediction result is estimated from the data of each pattern of the data of the initial condition of the numerical analysis of the meteorological phenomenon and/or the data of each pattern of the data of the analysis result of the meteorological phenomenon. Meteorological analysis inferiority estimation section,
A numerical analysis device, comprising: a power demand prediction unit that predicts a power demand based on the estimated degree of inferiority of the prediction result and numerical weather data. The numerical analysis device according to claim 4, wherein
From the data of the initial condition of the numerical analysis of the meteorological phenomenon and/or the data of each of the patterns of the numerical analysis result data of the meteorological phenomenon, there is an error in the meteorological numerical prediction that is the time evolution of the meteorological numerical analysis. Meteorological analysis inferiority estimation unit that estimates inferiority,
An initial condition increment processing unit that corrects the data of the initial condition based on the degree of inferiority of the weather analysis,
And a weather reanalysis processing unit that reanalyzes the meteorological phenomenon based on the corrected initial condition. A numerical analysis method executed in a numerical analysis device for performing numerical prediction,
A first step of classifying the data of the initial condition of the numerical analysis and/or the data of the analysis result of the numerical analysis into a plurality of patterns;
A second step of detecting the relationship between each of the patterns and the inferiority of the prediction result;
A third step of estimating the inferiority of the prediction result of the numerical prediction target based on the relationship with the inferiority of the prediction result detected for the pattern into which the numerical prediction target is classified. Characteristic numerical analysis method. The numerical analysis method according to claim 7, wherein
Inferiority data of the result of the first prediction regarding the state of the system targeted for the numerical prediction, and inferiority of the result of the second prediction based on the result of the prediction regarding the state of the system targeted for the numerical prediction. And a third step of displaying at least one of the data and the numerical analysis method. The numerical analysis method according to claim 8, wherein
In the second step,
The second data based on at least one of the data of each of the patterns of the initial condition data, the data of each pattern of the analysis result, and the inferiority data of the prediction result of each pattern. Prediction is performed and/or data of each of the patterns of the data of the initial condition, data of each of the patterns of the analysis result, data of inferiority of the prediction result of each of the patterns, and the second prediction A numerical analysis method, wherein the amount of preparation to be secured in case of a prediction failure is estimated based on at least one of the inferiority data included in. The numerical analysis method according to claim 7, wherein
The prediction target is a meteorological phenomenon,
In the first step,
Obtaining at least one of the data of the initial condition of the numerical analysis for the meteorological phenomenon and the data of the analysis result,
The pattern classification unit,
The numerical analysis method, wherein the acquired data of the initial condition of the numerical analysis for the meteorological phenomenon and/or the data of the analysis result is classified into a plurality of the patterns. The numerical analysis method according to claim 10, wherein
The inferiority of the prediction result is an error in the numerical weather prediction, which is the time evolution of the numerical weather analysis,
In the second step,
Estimating the inferiority of the prediction result from the data of each pattern of the data of the initial condition of the numerical analysis of the meteorological phenomenon and/or the data of each pattern of the data of the analysis result of the meteorological phenomenon. ,
A numerical analysis device, which predicts the power demand based on the estimated inferiority of the prediction result and numerical weather data. The numerical analysis method according to claim 10, wherein
In the second step,
From the data of the initial condition of the numerical analysis of the meteorological phenomenon and/or the data of each of the patterns of the numerical analysis result data of the meteorological phenomenon, there is an error in the meteorological numerical prediction that is the time evolution of the meteorological numerical analysis. Estimate the inferiority,
Based on the poorness of the meteorological analysis, amend the data of the initial conditions,
A numerical analysis method comprising reanalyzing the meteorological phenomenon based on the corrected initial condition.