Power generation quantity prediction device, power generation quantity prediction system, and program

Abstract

A power generation quantity prediction device according to one embodiment of the present invention is for predicting the power generation quantity by a power generation facility in a prescribed period in the future, and comprises: a past power generation quantity data acquisition unit that acquires past power generation quantity data, which is performance data regarding past power generation quantities in the power generation facility; a weather information acquisition unit that acquires predicted weather information in the prescribed period; a facility information acquisition unit that acquires facility information which includes at least positional information representing the facility position of the power generation facility; and a power generation quantity prediction unit that predicts the power generation quantity on the basis of a machine learning model into which predicted power generation quantities on the basis of the past power generation quantity data, the weather information, and a physical model have been input. The physical model predicts the power generation quantity on the basis of the weather information and the facility information.

Description

Power generation forecasting device, power generation forecasting system and program

[0001] The present invention relates to a power generation amount prediction device, a power generation amount prediction system, and a program.

[0002] In renewable energy power generation businesses, power generation volume fluctuates depending on weather conditions, so power generation forecasts are used to stabilize the power grid. Power generation companies create power generation plans for the power plants they own. If there is a discrepancy between the power generation plan and the actual results, costs will be incurred for the discrepancy. Therefore, highly accurate power generation forecasts are important to improve the profitability of power generation companies.

[0003] Patent Document 1 discloses a power generation prediction device that calculates a predicted value for power generation using a model constructed by machine learning using explanatory variables including weather forecast information for multiple meshes including the prediction point and a target variable corresponding to the amount of power generation from natural energy.

[0004] International Publication No. 2020 / 203854

[0005] Even if sophisticated modeling is performed in the estimation model for power generation, there is a limit to how much the prediction accuracy of power generation can be improved due to factors such as the influence of the installation conditions of the power generation equipment.

[0006] In order to solve the above-mentioned problems, an object of the power generation amount prediction device according to the present invention is to provide a power generation amount prediction device capable of highly accurate power generation amount prediction.

[0007] A power generation prediction device according to one embodiment of the present invention is a power generation prediction device that predicts the amount of power generated by a power generation facility for a predetermined future period, and includes a past power generation data acquisition unit that acquires past power generation data, which is actual data on the amount of power generated in the past at the power generation facility; a weather information acquisition unit that acquires weather information predicted for the predetermined period; an equipment information acquisition unit that acquires equipment information including at least location information indicating the installation location of the power generation facility; and a power generation prediction unit that predicts the amount of power generation based on a machine learning model to which the past power generation data, the weather information, and a physical model have been input, and the physical model predicts the amount of power generation based on the weather information and the equipment information.

[0008] According to the present invention, it is possible to predict the amount of power generation with high accuracy.

[0009] FIG. 1 is a block diagram illustrating an example of the overall configuration of a power generation prediction system according to a first embodiment of the present invention. FIG. 2 is a diagram illustrating the solar altitude angle and azimuth angle. FIG. 3 is a block diagram illustrating an example of the hardware configuration of a power generation prediction device according to a first embodiment of the present invention. FIG. 4 is a block diagram illustrating an example of the functional configuration of a power generation prediction device according to a first embodiment of the present invention. FIG. 5 is a flow diagram illustrating the power generation prediction process in the power generation prediction device according to the first embodiment of the present invention. FIG. 6 is a diagram illustrating the details of the processes in steps S104 to S106 of FIG. 5. FIG. 7A is a graph illustrating the actual power generation value, the predicted value by a machine learning model, and the predicted value by a physical model, with the vertical axis representing the power generation amount and the horizontal axis representing the date and time. FIG. 7B is a graph illustrating the MAE % value of the predicted power generation value relative to the actual power generation value. FIG. 8A is a graph illustrating the effect of using solar position information in the machine learning model of the power generation prediction device according to the first embodiment of the present invention when there is a decrease in power generation due to the shadow of a building. FIG. 8B is a graph illustrating the effect of using solar position information in the machine learning model of the power generation prediction device according to the first embodiment of the present invention when there is a decrease in power generation due to the shadow of a mountain. FIG. 9 is a block diagram showing an example of the overall configuration of a power generation amount prediction system according to a second embodiment of the present invention. FIG. 10 is a diagram showing an example of a power curve in a wind power generation facility. FIG. 11 is a block diagram showing an example of the functional configuration of a power generation amount prediction device according to a second embodiment of the present invention. FIG. 12 is a flow diagram showing the power generation amount prediction process in the power generation amount prediction device according to the second embodiment of the present invention. FIG. 13 is a diagram for explaining the details of the processes in steps S204 to S206 in FIG. 12. FIG. 14 is a diagram showing the results of power generation amount prediction in the power generation amount prediction device according to the second embodiment of the present invention. FIG. 15 is a configuration diagram of an energy supply system using the power generation amount prediction system according to this embodiment.

[0010] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. In this specification and drawings, components having substantially the same functional configurations are designated by the same reference numerals, and redundant description will be omitted.

[0011] [First embodiment] <Overall configuration> The overall configuration of a power generation amount prediction system 1 in a first embodiment will be described with reference to Fig. 1. Fig. 1 is a block diagram showing an example of the overall configuration of a power generation amount prediction system 1 according to a first embodiment of the present invention. The power generation amount prediction system 1 according to this embodiment is used as a power generation amount prediction system 1 that uses renewable energy, such as solar power generation, whose power generation amount depends on weather and natural conditions.

[0012] According to a method for estimating the power generation amount of a solar power generation system using an engineering technique that employs a physical model, it is difficult to eliminate the influence of, for example, the shadow of a building, which is caused by the installation conditions of the solar power generation equipment. Furthermore, due to the spatial representativeness of the weather forecast data used as input for the estimation model, there is a risk of bias with respect to the weather at the installation site of the solar power generation equipment or bias due to deterioration of the solar panels. Therefore, a power generation amount prediction system 1 according to the present embodiment has been developed that can solve these problems. Below, the power generation amount prediction system 1 will be described as an example of a system used in the case of solar power generation.

[0013] As shown in the figure, the power generation amount prediction system 1 includes a power generation amount prediction device 10 and a terminal device 20. The power generation amount prediction device 10 and the terminal device 20 are connected to each other so as to be able to communicate data with each other via a communication network N1 such as a local area network (LAN) or the Internet.

[0014] The power generation prediction system 1 is connected to an external data source 30 via a communication network N1 so as to be able to communicate data with the external data source 30. The external data source 30 includes a power generation management system 31, a weather information system 32, a facility information management system 33, and a solar position calculation system 34.

[0015] The external data source 30 is not limited to these and may include other data sources. For example, the external data source 30 may include other data sources that accumulate data acquired from the power generation amount management system 31, the weather information system 32, the facility information management system 33, and the solar position calculation system 34, and process and provide the data.

[0016] The power generation prediction device 10 is an information processing device such as a personal computer, workstation, or server that predicts the amount of power generated by a power generation facility for a predetermined future period. The power generation prediction device 10 acquires data necessary for prediction from an external data source 30 and generates a power generation prediction model. The power generation prediction device 10 transmits the power generation prediction result to a terminal device 20.

[0017] The terminal device 20 is an information processing terminal such as a personal computer, a smartphone, or a tablet terminal operated by a user of the power generation prediction system 1. The terminal device 20 receives the power generation prediction result from the power generation prediction device 10 and presents the prediction result to the user.

[0018] The power generation amount management system 31 transmits past power generation amount data, which is actual data on the amount of power generated in the power generation facility in the past, to the power generation amount prediction device 10. The predetermined period may be, for example, the past 10 years, but is not limited to this and may be any period. The power generation amount management system 31 can accumulate data on the amount of power generated at multiple locations.

[0019] The weather information system 32 is an information processing system that provides weather information. The weather information may include forecast data such as atmospheric pressure, wind speed, temperature, relative humidity, precipitation, cloud cover, and solar radiation. The weather information system 32 transmits weather forecast data for a power generation forecast date and time for a predetermined future period to the power generation prediction device 10. The predetermined period may be, for example, three days or one week, but is not limited to these, and may be various other periods.

[0020] The facility information management system 33 transmits information about the power generation facility to the power generation prediction device 10. When the power generation facility is a solar power generation facility, the information about the solar panel includes the elevation angle θ of the solar panel. １and azimuth angle φ １ The information on the solar panel may include information on the location of the facility, such as the latitude and longitude of the location where the solar panel is installed.

[0021] The solar position calculation system 34 is an information processing system that calculates solar position information. The solar position calculation system 34 calculates the solar position information by calculating the position of the sun based on, for example, the latitude and longitude of the location where the solar power generation facility is installed, as well as date and time information. Specifically, the solar position calculation system 34 calculates the solar position information by calculating the altitude angle θ of the sun relative to the solar panel. ２ and azimuth angle φ ２ The solar position calculation system 34 may be a device or a program that calculates solar position information.

[0022] 2, the solar altitude angle θ included in the solar position information ２ and azimuth angle φ ２ FIG. 2 shows the solar altitude angle θ ２ and azimuth angle φ ２ 1 is a diagram for explaining the altitude angle θ ２ can range from -90° to 90° with respect to the horizon. ２ can range from 0° to 360° with true north as the reference.

[0023] The overall configuration of the power generation prediction system 1 shown in FIG. 1 is one example, and various system configuration examples are possible depending on the application and purpose. For example, the power generation prediction system 1 may include multiple power generation prediction devices 10 and one or more terminal devices 20. For example, the power generation prediction device 10 may be realized by multiple computers, or may be realized as a cloud computing service. For example, the power generation prediction system 1 may be realized by a standalone computer. The division of devices such as the power generation prediction device 10 and the terminal device 20 shown in FIG. 1 is one example.

[0024] <Hardware Configuration> The hardware configuration of the power generation amount prediction device 10 in the first embodiment will be described with reference to Fig. 3. Fig. 3 is a block diagram showing an example of the hardware configuration of the power generation amount prediction device 10 according to the first embodiment of the present invention.

[0025] 3, the power generation prediction device 10 includes a central processing unit (CPU) 41, a read-only memory (ROM) 42, a random access memory (RAM) 43, a hard disk drive (HDD) 44, an input device 45, a display device 46, a communication interface (I / F) 47, and an external I / F 48. The CPU 41, the ROM 42, and the RAM 43 form a so-called computer. The hardware components of the power generation prediction device 10 are connected to each other via a bus line 49. The input device 45 and the display device 46 may be connected to the external I / F 48 for use.

[0026] The CPU 41 is a computing device that reads programs and data from a storage device such as the ROM 42 or the HDD 44 onto the RAM 43 and executes processing to realize overall control and functions of the power generation amount prediction device 10. The power generation amount prediction device 10 may have a GPU (Graphics Processing Unit) in addition to or instead of the CPU 41.

[0027] The ROM 42 is an example of a non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. The ROM 42 functions as a main storage device that stores various programs, data, etc. required for the CPU 41 to execute various programs installed in the HDD 44. Specifically, the ROM 42 stores boot programs such as a Basic Input / Output System (BIOS) and an Extensible Firmware Interface (EFI) that are executed when the power generation amount prediction device 10 is started, as well as data such as OS (Operating System) settings and network settings.

[0028] The RAM 43 is an example of a volatile semiconductor memory (storage device) in which programs and data are erased when the power is turned off. The RAM 43 is, for example, a dynamic random access memory (DRAM) or a static random access memory (SRAM). The RAM 43 provides a working area in which various programs installed in the HDD 44 are expanded when executed by the CPU 41.

[0029] The HDD 44 is an example of a non-volatile storage device that stores programs and data. The programs and data stored in the HDD 44 include an OS, which is basic software that controls the entire power generation prediction device 10, and applications that provide various functions on the OS. Note that the power generation prediction device 10 may use a storage device that uses flash memory as a storage medium (e.g., an SSD (Solid State Drive)) instead of the HDD 44.

[0030] The input device 45 includes a touch panel, operation keys and buttons, a keyboard and mouse, a microphone for inputting sound data such as voice, and the like, which are used by the user to input various signals.

[0031] The display device 46 is composed of a display such as a liquid crystal display or an organic EL (Electro-Luminescence) display for displaying a screen, a speaker for outputting sound data such as voice, and the like.

[0032] The communication I / F 47 is an interface that connects to a communication network and allows the power generation amount prediction device 10 to perform data communication.

[0033] The external I / F 48 is an interface with external devices, such as a drive device 50.

[0034] The drive device 50 is a device for loading a recording medium 51. The recording medium 51 here includes media that record information optically, electrically, or magnetically, such as CD-ROMs, flexible disks, and magneto-optical disks. The recording medium 51 may also include semiconductor memories that record information electrically, such as ROMs and flash memories. This allows the power generation amount prediction device 10 to read and / or write data from and to the recording medium 51 via the external I / F 48.

[0035] The various programs to be installed in the HDD 44 are installed, for example, by setting the distributed recording medium 51 in the drive device 50 connected to the external I / F 48 and reading the various programs recorded on the recording medium 51 by the drive device 50. Alternatively, the various programs to be installed in the HDD 44 may be installed by being downloaded via the communication I / F 47 from a network different from the communication network.

[0036] <Functional Configuration> Fig. 4 is a block diagram showing an example of the functional configuration of the power generation prediction device 10 according to the first embodiment of the present invention. As shown in the figure, the power generation prediction device 10 includes a past power generation data acquisition unit 11, a weather information acquisition unit 12, an equipment information acquisition unit 13, a solar position information acquisition unit 14, and a power generation prediction unit 15. The power generation prediction device 10 predicts the amount of power generated by a power generation facility for a predetermined future period. The predetermined period may be, for example, three days or one week, but is not limited to these, and may be various other periods.

[0037] The past power generation data acquiring unit 11 acquires past power generation data, which is actual data on the amount of power generated in the power generation facility in the past, from the power generation management system 31. The past power generation data may be, for example, data on the amount of power generated by a solar panel for a predetermined period of time.

[0038] The weather information acquisition unit 12 acquires weather information predicted for a predetermined period from the weather information system 32. The weather information may include atmospheric pressure, wind speed, temperature, relative temperature, precipitation, cloud cover, and solar radiation.

[0039] The facility information acquisition unit 13 acquires facility information including at least location information indicating the installation location of the power generation facility from the facility information management system 33. The facility information may be facility information related to the solar panels including at least one of the latitude and longitude of the location where the solar panels are installed, the angle of the solar panels, the power generation capacity of the solar panels, the capacity of the power conditioner, the conversion efficiency of the power conversion device, and the temperature coefficient of the solar panels.

[0040] The solar position information acquisition unit 14 acquires solar position information that indicates the position of the sun during a predetermined period, with respect to the installation position of the power generation facility. The solar position information is the altitude angle θ of the sun with respect to the installation position of the solar panel. ２ and azimuth angle φ ２ Includes.

[0041] The power generation amount prediction unit 15 may predict the power generation amount based on a machine learning model to which past power generation amount data and weather information are input. Alternatively, the power generation amount prediction unit 15 may predict the power generation amount based on a machine learning model to which solar position information, past power generation amount data, and weather information are input. Furthermore, the power generation amount prediction unit 15 may predict the power generation amount based on a physical model to which solar position information, weather information, and facility information are input.

[0042] When there is no past power generation data for the power generation facility, the power generation amount prediction unit 15 may determine the power generation amount predicted based on a physical model as the power generation amount for a predetermined period. When predicting the power generation amount based on a physical model, the power generation amount prediction unit 15 may make the prediction based on decomposition of the amount of solar radiation, calculation of incident light, estimation of the temperature of the cells included in the solar panel, estimation of the amount of DC power generation, and estimation of the amount of AC power generation.

[0043] The power generation amount prediction unit 15 may perform preprocessing on each acquired data. For example, when power generation amount data at a certain time is missing, the power generation amount prediction unit 15 may interpolate the data. For example, the data interpolation may be performed by interpolating the missing value with the most recent measurement value at that time.

[0044] The power generation amount prediction unit 15 may generate a machine learning model for predicting the power generation amount using the preprocessed data. In this case, the power generation amount prediction unit 15 generates a machine learning model corresponding to each combination of the teacher data and the prediction algorithm by learning each of the plurality of teacher data based on a predetermined prediction algorithm.

[0045] The training data is input to the machine model, and includes meteorological data such as atmospheric pressure, wind speed, temperature, relative humidity, precipitation, cloud cover, and solar radiation, as well as the solar altitude angle θ relative to the solar panel. ２ and azimuth angle φ ２ The weather information input to the machine learning model includes at least one of atmospheric pressure, wind speed, temperature, relative humidity, precipitation, cloud cover, and solar radiation.

[0046] The power generation amount prediction unit 15 may further generate a plurality of machine learning models with different explanatory variables for a combination of teacher data and a prediction algorithm. The explanatory variables may be determined in advance as explanatory variables that can be used depending on the prediction algorithm.

[0047] The machine learning model is a model for predicting power generation amount using a statistical method based on a target variable and an explanatory variable. The machine learning model includes at least a prediction algorithm for predicting power generation amount. Examples of the machine learning model may include a random forest, a LightGBM (Light Gradient Boosting Machine), or an XGBoost (eXtreme Gradient Boosting).

[0048] The power generation amount prediction unit 15 predicts the power generation amount using a physical model based on weather information and facility information, and calculates a physical model predicted value. The physical model predicted value may be, for example, a power generation amount prediction every 10 minutes, 30 minutes, etc., but is not limited to this. The information about the solar panel includes information about the solar panel, such as the elevation angle and azimuth angle of the solar panel, the capacity and conversion efficiency of the power conditioner, the temperature coefficient of the solar panel, and location information such as the latitude and longitude of the location where the solar panel is installed. The weather information input to the physical model includes at least one of wind speed, temperature, and solar radiation.

[0049] The power generation amount prediction unit 15 can decompose the amount of solar radiation, calculate incident light, estimate the temperature of the cells included in the solar panel, estimate the amount of DC power generation, and estimate the amount of AC power generation. The amount of solar radiation is decomposed by decomposing global horizontal irradiance (GHI) into diffuse horizontal irradiance (DHI) and direct normal irradiance (DNI). The calculation of incident light is performed taking into account the wavelength dependence of reflection on the surface of the solar panel and atmospheric absorption.

[0050] The temperature of the cells in a solar panel is estimated mainly based on the position of the sun, the amount of solar radiation, the temperature, and the wind speed. The DC power generation amount is estimated by converting it into the DC power generation amount in the solar panel based on the results of decomposing the amount of solar radiation, calculating the incident light, and estimating the cell temperature. The AC power generation amount is estimated by converting it into the AC power generation amount using an inverter model based on the estimated DC power generation amount.

[0051] The physical model predicted value predicted by the power generation amount prediction unit 15 does not need to use data on the amount of power generated by the solar panels for a predetermined period of time. In this case, power generation amount prediction is possible even when there is no data on the amount of power generated in the past, for example, in the location where a solar panel is to be newly installed.

[0052] The power generation amount prediction unit 15 predicts the power generation amount using the power generation amount prediction based on the generated machine learning model and the physical model. Therefore, the power generation amount prediction unit 15 can predict the power generation amount while eliminating the influence of, for example, the shadow of a building, which is caused by the installation conditions of the solar panel. Furthermore, the power generation amount prediction unit 15 can predict the power generation amount while eliminating the bias with respect to the weather at the installation site of the solar panel caused by the spatial representativeness of the weather information and the bias due to the deterioration of the solar panel. Note that the power generation amount prediction by the power generation amount prediction unit 15 may be, for example, a prediction every 30 minutes, but is not limited to this.

[0053] The power generation amount prediction unit 15 may perform a predetermined process to convert the predicted power generation amount into outputtable data and then output the data.

[0054] <Processing Procedure> Here, an inference process depending on whether or not data on past power generation amounts is available will be described. Fig. 5 is a flow chart showing the power generation amount prediction process in the power generation amount prediction device 10 according to the first embodiment of the present invention.

[0055] First, the power generation prediction device 10 acquires weather information for a predetermined future period to be predicted, facility information including the location information of the power generation facility, and solar position information (step S101). Next, the power generation prediction unit 15 predicts the power generation amount based on the physical model to which the acquired information is input (step S102).

[0056] If a machine learning model is available for the power generation facility to be predicted (No in step S103), the power generation prediction unit 15 acquires a predicted power generation value based on a physical model (step S104).Then, the power generation prediction unit 15 predicts the power generation amount based on the machine learning model (step S105).Furthermore, the power generation prediction unit 15 outputs the predicted power generation amount data (step S106).

[0057] If there is no machine learning model for the solar panel to be predicted (Yes in step S103), the power generation amount prediction unit 15 outputs power generation amount data predicted based on a physical model (step S107). Note that a case where there is no machine learning model means, for example, a case where there is no past power generation amount data, such as a location where power generation equipment has been newly installed.

[0058] 6 is a diagram for explaining the details of the processing of steps S104 to S106 in FIG. 5 . As shown in the figure, past power generation amount data, weather information, solar position information, and a physics model are input to the machine learning model. Weather information and facility information are input to the physics model. The weather information input to the machine learning model is, for example, atmospheric pressure, wind speed, temperature, relative temperature, precipitation, cloud cover, and solar radiation, while the weather information input to the physics model is, for example, wind speed, temperature, and solar radiation. In addition, the facility information input to the physics model is, for example, the latitude and longitude of the locations where the solar panels are installed, the angle of the solar panels, the capacity of a power conversion device included in the solar power generation facility, and the conversion efficiency and temperature coefficient of the power conversion device.

[0059] Forecasting power generation using machine learning models requires past power generation data. Furthermore, machine learning models predict power generation based on solar position information, allowing for power generation predictions that take the sun's position into account. This eliminates the influence of shadows from mountains and buildings, for example, and increases the accuracy of predictions. Physical models do not require past power generation data, so power generation predictions can be made even if past power generation data is unavailable, such as when a new solar power generation facility is installed.

[0060] <Prediction Accuracy> The prediction accuracy of the power generation prediction device 10 in this embodiment will be described with reference to Figs. 7 and 8. In the following, the MAE % value is used as an evaluation index for the prediction error. The MAE % value is the mean absolute error normalized by the maximum possible power generation (kWh) of the photovoltaic power generation facility. The maximum possible power generation is the larger of the power generation capacity of the solar panel and the capacity of the power conditioner.

[0061] The MAE % value is calculated by the following formula (1): where i is the 30-minute frame number, n is the predicted frame number, E ｏｂｓ is the actual power generation amount, E ｐｒｅｄ is the power generation forecast, E ｍａｘ is the maximum amount of power that can be generated by the solar power generation facility. Note that the smaller the MAE % value, the higher the prediction accuracy.

[0062]

[0063] 7A and 7B are diagrams showing the results of power generation prediction by the power generation prediction device 10 according to the first embodiment of the present invention. Fig. 7A shows a graph of the actual power generation amount, the predicted value by the machine learning model, and the predicted value by the physical model, with the vertical axis representing the power generation amount and the horizontal axis representing the date and time. Fig. 7B shows the MAE % value of the predicted power generation amount relative to the actual power generation amount. As shown in Fig. 7A, it can be seen that the predicted value by the machine learning model tends to be closer to the actual value than the predicted value by the physical model.

[0064] 7B, the MAE% of the predicted power generation amount relative to the actual power generation amount is 6.1% for the full year for the physical model prediction, while the MAE% of the machine learning model prediction is 4.9%, and the MAE% of the machine learning model prediction is smaller than the physical model prediction in every month. This also shows that the machine learning model predictions tend to be closer to the actual values ​​than the physical model predictions.

[0065] 8A and 8B are graphs showing the effect of using solar position information in the machine learning model in the power generation prediction device 10 according to the first embodiment of the present invention. The vertical axis represents the amount of power generated, and the horizontal axis represents the date and time. The graphs show the actual value of the amount of power generated, the predicted value using a physical model, the predicted value using a machine learning model that does not use solar position information, and the predicted value using a machine learning model that uses solar position information. Fig. 8A shows the case where the amount of power generated is reduced due to the shadow of a building, and Fig. 8B shows the case where the amount of power generated is reduced due to the shadow of a mountain. The vertical axis represents the amount of power generated, and the horizontal axis represents the passage of time.

[0066] The data shown in Figure 8A is for two days, with the largest amount of power generation occurring during the daytime. As shown in the figure, the power generation amount predicted by the machine learning model is closer to the actual value than the power generation amount predicted by the physical model.

[0067] 8A, the MAE % of the predicted power generation amount relative to the actual power generation amount was 7.3% for the physical model prediction, 5.1% for the machine learning model without using solar position information, and 4.8% for the machine learning model with solar position information. This shows that the accuracy of power generation prediction is improved by using the machine learning model with solar position information.

[0068] FIG. 8B shows data for half a day. As it approaches noon, the amount of power generated increases. In the morning, as indicated by the arrow in the figure, the amount of power generated is low due to the influence of the mountain shadow. As shown in the figure, it can be seen that the power generation amount predicted by the machine learning model is closer to the actual value than the power generation amount predicted by the physical model. In addition, as shown in FIG. 8B, even when the machine learning model is used, the power generation amount predicted using solar position information is closer to the actual value than when solar position information is not used.

[0069] <Effects of the embodiment> The power generation prediction device 10 according to this embodiment predicts the amount of power generation based on a machine learning model to which past power generation data, weather information, and solar position information are input. Furthermore, since the power generation prediction device 10 predicts the amount of power generation based on a machine learning model to which a physical model prediction value is input, it is possible to remove bias in the prediction due to weather data and the environment, such as deterioration of the power generation equipment. Therefore, the power generation prediction device 10 according to this embodiment can predict the amount of power generation with high accuracy.

[0070] Second Embodiment <Overall Configuration> The overall configuration of a power generation amount prediction system 1 in a second embodiment will be described with reference to Fig. 9. Fig. 9 is a block diagram showing an example of the overall configuration of a power generation amount prediction system 1 according to a second embodiment of the present invention. The power generation amount prediction system 1 according to this embodiment will be described as an example of a system used in the case of wind power generation. Note that components that are the same as those already described in the first embodiment will be assigned the same reference numerals, and duplicated descriptions will be omitted.

[0071] The power generation prediction system 1 is connected to an external data source 30 via a communication network N1 so as to be able to communicate data with the external data source 30. The external data source 30 includes a power generation prediction system 31, a weather information system 32, and a facility information management system 33. Unlike the first embodiment, the power generation prediction system 1 according to this embodiment does not include a solar position calculation system 34.

[0072] The facility information management system 33 transmits information about the power generation facility to the power generation amount prediction device 10. When the power generation facility is a wind power generation facility, the information about the power generation facility may include wind turbine facility information, which is information including the relationship between wind speed and power output indicated by a power curve in the wind power generation facility, as shown in Fig. 10. The power output may be either the amount of power generation or the power output.

[0073] 11 is a block diagram showing an example of the functional configuration of a power generation amount prediction device 10 according to the second embodiment of the present invention. As shown in the figure, the power generation amount prediction device 10 includes a past power generation amount data acquisition unit 11, a weather information acquisition unit 12, a facility information acquisition unit 13, and a power generation amount prediction unit 15.

[0074] The facility information acquisition unit 13 acquires wind turbine facility information, which is information including at least the relationship between wind speed and power output indicated by a power curve in the wind power generation facility, from the facility information management system 33 .

[0075] The power generation amount prediction unit 15 may predict the power generation amount based on a machine learning model to which past power generation amount data and weather information are input. Alternatively, the power generation amount prediction unit 15 may predict the power generation amount based on a machine learning model to which wind turbine facility information, past power generation amount data, and weather information are input.

[0076] The power generation amount prediction unit 15 may perform preprocessing on each acquired data. For example, when power generation amount data at a certain time is missing, the power generation amount prediction unit 15 may interpolate the data. For example, the data interpolation may be performed by interpolating the missing value with the most recent measurement value at that time.

[0077] The power generation amount prediction unit 15 predicts the power generation amount using a physical model based on meteorological information and facility information, and calculates a physical model prediction value. In the physical model prediction method according to this embodiment, wind speed is input, and the power generation amount estimated based on a power curve such as that shown in Fig. 9 is output. The physical model prediction value may be, but is not limited to, a power generation amount prediction every 10 minutes or 30 minutes, for example.

[0078] <Processing Procedure> Here, an inference process depending on whether or not data on past power generation amounts is available will be described. Fig. 12 is a flowchart showing the power generation amount prediction process in the power generation amount prediction device 10 according to the second embodiment of the present invention.

[0079] First, the power generation prediction device 10 acquires weather information for a predetermined future period to be predicted and facility information including location information of the power generation facility (step S201). Next, the power generation prediction unit 15 predicts the power generation amount based on the physical model to which the acquired information is input (step S202).

[0080] If a machine learning model is available for the power generation facility to be predicted (No in step S203), the power generation prediction unit 15 acquires a predicted power generation value based on a physical model (step S204).Then, the power generation prediction unit 15 predicts the power generation amount based on the machine learning model (step S205).Furthermore, the power generation prediction unit 15 outputs the predicted power generation amount data (step S206).

[0081] If there is no machine learning model for the wind turbine equipment information to be predicted (Yes in step S203), the power generation amount prediction unit 15 outputs power generation amount data predicted based on a physical model (step S207). Note that a case where there is no machine learning model means, for example, that there is no past power generation amount data, such as in a location where power generation equipment has been newly installed.

[0082] Fig. 13 is a diagram for explaining the details of the processing of steps S204 to S206 in Fig. 12. As shown in the figure, past power generation amount data, weather information, and a physics model are input to the machine learning model. Weather information and facility information are input to the physics model. The weather information input to the machine learning model is, for example, atmospheric pressure, wind speed, air temperature, relative temperature, precipitation, cloud cover, and solar radiation, and the weather information input to the physics model is, for example, wind speed. Furthermore, the facility information input to the physics model is, for example, wind turbine facility information, and may include a power curve.

[0083] While machine learning models require historical power generation data to predict power generation, physical models do not require such data, making it possible to predict power generation even if no historical power generation data is available, such as when a new wind power generation facility is installed.

[0084] <Prediction Accuracy> The prediction accuracy of the power generation amount prediction device 10 in this embodiment will be described with reference to Fig. 14. Note that, in the following, as in the description of the first embodiment, the MAE % value is used as an evaluation index for the prediction error.

[0085] 14 is a diagram showing the results of power generation prediction by the power generation prediction device 10 according to the second embodiment of the present invention. In FIG. 14, the vertical axis represents the power generation amount, and the horizontal axis represents the date and time. The graph shows the actual power generation amount, the predicted value by the machine learning model, and the predicted value by the physical model. As shown in the figure, the predicted value by the machine learning model tends to be closer to the actual value than the predicted value by the physical model.

[0086] [Other Embodiments] The power generation prediction device 10 according to this embodiment is used as a power generation prediction device in power generation projects using renewable energy, such as solar power generation, wind power generation, and hydroelectric power generation, where the amount of power generated is dependent on weather and natural conditions. The power generation prediction device 10 can also be used in systems that control energy resources, such as power generation facilities using various natural energies, large-scale storage batteries that can supply stored power, and hydrogen power generation, and adjust the balance between power supply and demand. In addition to these, the power generation prediction device 10 according to this embodiment also has an embodiment as a power generation prediction method. A specific example will be described below.

[0087] FIG. 15 is a configuration diagram of an energy supply system 100 using the power generation prediction system 1 according to this embodiment. The energy supply system 100 shown in FIG. 15 includes an electric utility system 2 including the power generation prediction system 1, a consumer information processing terminal 3, and consumer energy resources 4. The electric utility system 2 includes the power generation prediction system 1, a virtual power plant system 60, a supply and demand management system 61, a power transmission and distribution system 62, a centralized power source 63, a distributed power source 64, a storage battery 65, and a hydrogen power generation facility 66. The supply and demand management system 61, the power transmission and distribution system 62, the centralized power source 63, the distributed power source 64, the storage battery 65, and the hydrogen power generation facility 66 are owned, for example, by an electric utility of the electric utility system 2. The consumer energy resources 4 shown in FIG. 15 include a solar power generation system 71, a wind power generation system 72, a storage battery system 73, an electric vehicle 74, a vehicle-to-home (V2H) device 75, and a hydrogen production device 76. Note that the electric utility may own some or all of the consumer energy resources 4.

[0088] The virtual power plant system 60 utilizes information and communication technologies such as the Internet of Things (IoT) to remotely control the charging and discharging of consumer energy resources 4 via a communication network such as the Internet. The virtual power plant system 60 controls the aggregation of surplus power from consumer energy resources 4 connected to the electric utility system 2 by remotely controlling consumer energy resources 4, centralized power sources 63, distributed power sources 64, storage batteries 65, and hydrogen power generation facilities 66. Surplus power is power generated by or stored in consumer energy resources 4 that exceeds the power consumed by the consumer.

[0089] The virtual power plant system 60 determines the supply destination of surplus electricity collected from the energy resources 4 of the consumers, the centralized power source 63, the distributed power source 64, the storage battery 65, and the hydrogen power generation facility 66. The virtual power plant system 60 can control the supply of surplus electricity to other consumers, and has the adjustment power to adjust the supply and demand balance.

[0090] The virtual power plant system 60 can return the profits gained from supplying surplus power to the selected supply destination to the consumers. For example, the virtual power plant system 60 may use the profits gained from supplying surplus power to the selected supply destination to reduce electricity charges, award points, or provide new services.

[0091] The information processing terminal 3 of the consumer is a computer operated by the consumer, such as a personal computer (PC), a home energy management system (HEMS), or a smartphone. The information processing terminal 3 of the consumer displays information received from the virtual power plant system 60.

[0092] The supply and demand management system 61 manages the supply and demand of electricity for each consumer in order to achieve overall optimization of the electricity supply and demand based on the amount of power generation predicted by the power generation prediction system 1. The electricity transmission and distribution system 62 is, for example, a system of a wholesale electricity market, a capacity market, or a supply and demand adjustment market. The wholesale electricity market is a trading market where electricity volume is bought and sold. The capacity market is a trading market where future supply capacity is bought and sold. The supply and demand adjustment market is a trading market where adjustment capacity is bought and sold.

[0093] The centralized power source 63 is a large-scale power generation facility installed at a location far from the power demand area, such as a thermal power generation facility, a hydroelectric power generation facility, or a nuclear power generation facility. The distributed power source 64 is a relatively small-scale power generation facility installed in a dispersed manner near the power demand area, such as a solar power generation facility, a wind power generation facility, or another renewable energy power generation facility. The storage battery 65 is a power storage facility that stores energy for later use or for use as an energy regulator. The hydrogen power generation facility 66 generates electricity using hydrogen supplied from, for example, a hydrogen cartridge as an energy source and supplies the generated electricity to devices, etc. If surplus electricity using renewable energy is available, hydrogen produced from the surplus electricity can be stored in the hydrogen power generation device 66. In the hydrogen power generation facility 66, electricity is generated using hydrogen produced and stored by the hydrogen production device 76 using the surplus electricity.

[0094] The centralized power source 63, the distributed power source 64, the storage battery 65, and the hydrogen power generation facility 66 are examples of multiple energy resources. The virtual power plant system 60 controls the power supply and demand of an energy supply platform that integrates multiple energy resources. The energy supply platform, for which the virtual power plant system 60 controls the power supply and demand, is connected to a solar power generation system 71, a wind power generation system 72, a storage battery system 73, an electric vehicle 74, a V2H device 75, and a hydrogen production device 76, which are examples of energy resources 4 of consumers.

[0095] A solar power generation system 71 and a wind power generation system 72, which are examples of the consumer's energy resources 4, are examples of renewable energy power generation facilities installed at the consumer's facility. A storage battery system 73 is an example of power storage facilities installed in a home or the like. A V2H device 75 is a system that uses an on-board storage battery installed in an electric vehicle 74 as a home power source.

[0096] 15 is merely an example, and it goes without saying that there are various system configuration examples depending on the application and purpose. For example, the virtual power plant system 60 can have various configurations, such as a configuration in which multiple systems work together to perform processing.

[0097] [Supplementary Information] Each function of the above-described embodiment can be realized by one or more processing circuits included in the power generation amount prediction device 10. Here, the term "processing circuit" in this specification includes a processor programmed to execute each function by software, such as a CPU (Central Processing Unit) or GPU (Graphics Processing Unit) implemented by an electronic circuit, as well as devices such as an ASIC (Application Specific Integrated Circuit), DSP (Digital Signal Processor), FPGA (Field Programmable Gate Array), and conventional circuit modules designed to execute each function described above.

[0098] Summary of Embodiments This specification discloses at least the power generation prediction device, power generation prediction system, power generation prediction method, and program described below. (Supplementary Note 1) An power generation prediction device that predicts the amount of power generated by a power generation facility for a predetermined future period, comprising: a past power generation data acquisition unit that acquires past power generation data, which is actual data on the amount of power generated in the past at the power generation facility; a weather information acquisition unit that acquires weather information predicted for the predetermined period; an equipment information acquisition unit that acquires equipment information including at least location information indicating the installation location of the power generation facility; and a power generation prediction unit that predicts the amount of power generated based on a machine learning model to which the past power generation data, the weather information, and a physical model have been input, wherein the physical model predicts the amount of power generated based on the weather information and the equipment information. (Supplementary Note 2) The power generation prediction device described in Supplementary Note 1, wherein the power generation facility is a photovoltaic power generation facility including solar panels, and further comprises a solar position information acquisition unit that acquires solar position information indicating the position of the sun during the predetermined period based on the installation location of the photovoltaic power generation facility, and wherein the solar position information is further input to the machine learning model. (Supplementary Note 3) The power generation prediction device according to Supplementary Note 2, wherein the solar position information includes an altitude angle and an azimuth angle of the sun with respect to an installation position of the solar power generation facility. (Supplementary Note 4) The power generation prediction device according to Supplementary Note 2 or 3, wherein the facility information includes the facility information related to solar panels, and the facility information related to the solar panels includes at least one of the latitude and longitude of a location where the solar panels are installed, an angle of the solar panels, a capacity of a power conditioner included in the solar power generation facility, and a conversion efficiency and a temperature coefficient of the power conditioner. (Supplementary Note 5) The power generation prediction device according to any one of Supplements 1 to 4, wherein the weather information input to the machine learning model includes at least one of atmospheric pressure, wind speed, temperature, relative temperature, precipitation, cloud cover, and solar radiation. (Supplementary Note 6) The power generation prediction device according to any one of Supplements 1 to 5, wherein the weather information input to the physical model includes at least one of wind speed, temperature, and solar radiation.(Supplementary Note 7) The power generation prediction device according to any one of Supplementary Notes 1 to 6, wherein the power generation amount prediction unit determines the power generation amount predicted based on the physical model as the power generation amount for the predetermined period when there is no past power generation amount data for the power generation facility. (Supplementary Note 8) The power generation prediction device according to Supplementary Note 7, wherein, when predicting the power generation amount based on the physical model, the power generation amount prediction unit makes the prediction based on decomposition of solar radiation, calculation of incident light, estimation of temperatures of cells included in a solar panel, estimation of DC power generation amount, and estimation of AC power generation amount. (Supplementary Note 9) The power generation prediction device according to Supplementary Note 1, wherein the power generation facility is a wind power generation facility including a wind turbine facility, and the facility information acquisition unit acquires wind turbine facility information that is information indicating at least a relationship between wind power and power generation output of the wind turbine facility. (Supplementary Note 10) The power generation prediction device according to Supplementary Note 9, wherein the weather information input to the physical model includes at least wind speed. (Supplementary Note 11) A power generation prediction system that predicts the amount of power generated by a power generation facility for a predetermined future period, comprising: a past power generation data acquisition means that acquires past power generation data, which is actual data on the amount of power generated in the past at the power generation facility; a weather information acquisition means that acquires weather information predicted for the predetermined period; a facility information acquisition means that acquires facility information including at least location information indicating the installation location of the power generation facility; and a power generation prediction means that predicts the amount of power generated based on a machine learning model to which the past power generation data, the weather information, and a physical model have been input, wherein the physical model predicts the amount of power generated based on the weather information and the facility information.and a machine learning model into which the amount of power generated predicted based on the past power generation amount data, the weather information, and a physical model is input, wherein the physical model predicts the amount of power generated based on the past power generation amount data, the weather information, and a physical model that predicts the amount of power generated based on the past power generation amount data, the physical model predicts the amount of power generated based on the weather information and the equipment information. (Supplementary Note 12) A power generation prediction method executed by a power generation prediction device that predicts the amount of power generated by a power generation facility for a predetermined period in the future, the method comprising: acquiring past power generation amount data, which is actual data of the amount of power generated in the past at the power generation facility; acquiring weather information predicted for the predetermined period in the future; acquiring facility information including at least location information indicating the installation location of the power generation facility; and predicting the amount of power generated based on the machine learning model into which the amount of power generated predicted based on the past power generation amount data, the weather information, and a physical model that predicts the amount of power generated based on the weather information and the equipment information. (Supplementary Note 13) A non-transitory computer-readable medium that stores a program that causes a computer to execute the following processes.

[0099] Although the embodiments of the present invention have been described in detail above, the present invention is not limited to these embodiments, and various modifications and changes are possible within the scope of the gist of the present invention described in the claims.

[0100] This application claims priority from Japanese Patent Application No. 2024-54987, filed with the Japan Patent Office on March 28, 2024, the entire contents of which are incorporated herein by reference.

[0101] 1: Power generation amount prediction system 10: Power generation amount prediction device 11: Past power generation amount data acquisition unit 12: Weather information acquisition unit 13: Equipment information acquisition unit 14: Solar position information acquisition unit 15: Power generation amount prediction unit

Claims

1. A power generation prediction device that predicts the amount of power generated by a power generation facility for a predetermined future period, comprising: a past power generation data acquisition unit that acquires past power generation data, which is actual data on the amount of power generated in the past at the power generation facility; a weather information acquisition unit that acquires weather information predicted for the predetermined period; a facility information acquisition unit that acquires facility information including at least location information indicating the installation location of the power generation facility; and a power generation prediction unit that predicts the amount of power generated based on a machine learning model to which the past power generation data, the weather information, and a physical model have been input, wherein the physical model predicts the amount of power generated based on the weather information and the facility information.

2. The power generation prediction device according to claim 1, wherein the power generation facility is a solar power generation facility including solar panels, and further comprises a solar position information acquisition unit that acquires solar position information indicating the position of the sun during the specified period based on the installation position of the solar power generation facility, and the solar position information is further input into the machine learning model.

3. The power generation prediction device according to claim 2, wherein the solar position information includes the solar altitude angle and azimuth angle based on the installation position of the solar power generation facility.

4. The power generation prediction device of claim 2, wherein the equipment information includes equipment information related to solar panels, and the equipment information related to the solar panels includes at least one of the latitude and longitude of the location where the solar panels are installed, the angle of the solar panels, the capacity of a power conditioner included in the solar power generation equipment, and the conversion efficiency and temperature coefficient of the power conditioner.

5. The power generation prediction device according to claim 1, wherein the weather information input to the machine learning model includes at least one of atmospheric pressure, wind speed, temperature, relative temperature, precipitation, cloud cover, and solar radiation.

6. The power generation prediction device according to claim 1, wherein the meteorological information input to the physical model includes at least one of wind speed, temperature, and solar radiation.

7. The power generation prediction device according to claim 1, wherein, when there is no past power generation data for the power generation facility, the power generation prediction unit determines the power generation amount predicted based on the physical model as the power generation amount for the specified period.

8. The power generation prediction device according to claim 7, wherein the power generation prediction unit, when predicting the power generation amount based on the physical model, makes the prediction based on decomposition of solar radiation, calculation of incident light, estimation of the temperature of cells included in the solar panel, estimation of DC power generation amount, and estimation of AC power generation amount.

9. The power generation prediction device according to claim 1, wherein the power generation facility is a wind power generation facility including a wind turbine facility, and the facility information acquisition unit acquires wind turbine facility information, which is information indicating the relationship between wind power and power generation output of at least the wind turbine facility.

10. The power generation prediction device according to claim 9, wherein the meteorological information input to the physical model includes at least wind speed.

11. A power generation prediction system that predicts the amount of power generated by a power generation facility for a specified future period, comprising: a past power generation data acquisition means that acquires past power generation data, which is actual data on the amount of power generated in the past at the power generation facility; a weather information acquisition means that acquires weather information predicted for the specified period; a facility information acquisition means that acquires facility information including at least location information indicating the installation location of the power generation facility; and a power generation prediction means that predicts the amount of power generated based on a machine learning model to which the past power generation data, the weather information, and a physical model have been input, wherein the physical model predicts the amount of power generated based on the weather information and the facility information.

12. A non-transitory computer-readable medium storing a program that causes a computer to execute the following processes: acquiring past power generation data, which is actual data on the amount of power generated in a power generation facility in the past; acquiring weather information predicted for a predetermined period in the future; acquiring facility information including at least location information indicating the installation location of the power generation facility; and predicting power generation based on a machine learning model to which the past power generation data, the weather information, and a power generation amount predicted based on a physical model that predicts the power generation amount based on the weather information and the facility information are input.

Images