Information processing device, information processing method, and program

Abstract

Even when there are gaps in the actual data on renewable energy generation, the amount of electricity generated can be estimated with high accuracy. [Solution] The information processing device of the embodiment includes a processing unit. The processing unit acquires weather data and actual power generation values ​​measured by a power generation sensor, generates a supplemented power generation amount by supplementing the missing actual power generation values ​​based on the weather data and the actual power generation values, and estimates the power generation amount of renewable energy sources based on the supplemented power generation amount, the weather data and a power generation estimation model.

Description

Technical Field

[0001] Embodiments of the present invention relate to an information processing apparatus, an information processing method, and a program.

Background Art

[0002] Renewable energy (RE) power sources such as solar power generation devices and wind power generation devices have power generation amounts that vary depending on weather conditions, so it is necessary to estimate the power generation amount at the time of supply planning or the like. For example, a model for estimating the power generation amount at an estimation point is constructed by machine learning from weather estimation information of a mesh including the estimation point and a plurality of meshes around the estimation point, and the power generation amount is estimated using this model. A method is known.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, in the above-described estimation method, it is not considered that there are missing values in the actual power generation amounts. Further, when such missing values are supplemented with the average value of the actual values or the data before and after the missing values, there is a problem that the estimation accuracy deteriorates because they are different from the original actual values.

[0005] The problem to be solved by the present invention is to accurately estimate the power generation amount even when there are missing values in the actual power generation amount of a renewable energy power source.

Means for Solving the Problems

[0006] The information processing device of this embodiment includes a processing unit. The processing unit acquires weather data and actual power generation values ​​measured by a power generation sensor, generates a supplemented power generation amount by supplementing the missing values ​​in the actual power generation value based on the weather data and the actual power generation value, and estimates the power generation amount of renewable energy sources based on the supplemented power generation amount, the weather data, and a power generation estimation model. [Brief explanation of the drawing]

[0007] [Figure 1] Hardware configuration diagram of the information processing device of the first embodiment. [Figure 2] Functional diagram of the processing unit of the first embodiment. [Figure 3] A diagram showing an example of actual power generation figures. [Figure 4] A diagram illustrating an example of weather data. [Figure 5] A diagram showing an example of extracted power generation figures and weather data. [Figure 6] A diagram showing the relationship between wind speed and power generation. [Figure 7] A diagram showing an example of input data for a power generation estimation model. [Figure 8] A flowchart of the power generation estimation process in the first embodiment. [Figure 9] Functional diagram of the processing unit of the first embodiment. [Figure 10] A diagram showing an example of actual weather data during the learning period. [Figure 11] A diagram illustrating an example of outlier detection information and missing value detection information during the learning period. [Figure 12] A diagram illustrating an example of supplementary information used during the learning period. [Figure 13] A diagram showing an example of actual power generation figures during the learning period. [Figure 14] A diagram showing an example of the label data used to supplement the learning period. [Figure 15] Flowchart of the learning process in the second embodiment. [Figure 16] Flowchart of the power generation estimation process in the second embodiment. [Modes for carrying out the invention]

[0008] Hereinafter, embodiments of the information processing apparatus, information processing method, and program of the present invention will be described while referring to the attached drawings.

[0009] (First Embodiment)

[0010] FIG. 1 is a diagram showing an example of the hardware configuration of the information processing apparatus 1 according to the present embodiment. The information processing apparatus 1 is a computer device and includes an input unit 2, a display unit 3, a storage unit 4, a communication unit 5, and a processing unit 6. Further, each of the units 2 to 6 is communicably connected to each other by a bus 7.

[0011] The input unit 2 is an input device such as a mouse, a keyboard, or a touch panel. The input unit 2 receives information corresponding to an operation by the user as an instruction signal and outputs the instruction signal to the processing unit 6.

[0012] The display unit 3 is a display device such as an LCD (Liquid Crystal Display). The display unit 3 displays various information based on a display signal from the processing unit 6.

[0013] The storage unit 4 is a storage medium capable of storing information by a semiconductor such as a flash memory, or a storage medium capable of storing information magnetically or optically.

[0014] The communication unit 5 communicates information with an external device via a network in response to control from the processing unit 6.

[0015] The processing unit 6 is composed of, for example, a CPU (Central Processing Unit), a RAM (Random Access Memory), and a ROM (Read Only Memory). The CPU executes various processes based on programs stored in the ROM, the storage unit 4, etc., using a predetermined area of the RAM as a work area.

[0016] Figure 2 shows an example of the functional configuration of the processing unit 6 according to this embodiment. The processing unit 6 has, as functional units, a power generation amount supplementation data acquisition unit 601, a supplemented power generation amount generation unit 602, a weather data acquisition unit 603, a power generation amount estimation unit 604, and an output unit 605.

[0017] Furthermore, some or all of the functional units 601 to 605 may be implemented by dedicated hardware such as an integrated circuit (IC). In addition, in the following, processing other than that performed by the functional units 601 to 605, or processing performed by multiple of the functional units 601 to 605, may be described with the processing unit 6 as the main operator.

[0018] The power generation data acquisition unit 601 acquires actual power generation values ​​(actual power generation values) from the power generation DB (Data Base) and weather data from the weather data DB. Here, the power generation DB and the weather data DB are databases built on, for example, an external server device of the information processing device 1.

[0019] The actual power generation data stored in the power generation database is, for example, data measured by power generation sensors installed in power plants with renewable energy sources. The power plants may be wind power plants, solar power plants, or other types of renewable energy power plants. The weather data stored in the weather data database is, for example, NWP (Numerical Weather Prediction) data. NWP data is data such as wind speed, wind direction, and temperature obtained by estimating changes in the Earth's atmosphere, etc., through numerical simulations.

[0020] The power generation amount supplementation data acquisition unit 601 extracts actual power generation values ​​and weather data for a predetermined range from the acquired data based on the supplementation data range specification parameter, and transmits them to the supplemented power generation amount generation unit 602. Here, the supplementation data range specification parameter is a parameter for specifying data for a predetermined range up to the latest time period (for example, 10 days' worth). The latest time period is the latest time period among the data that can be acquired (for example, if the length of the time period is 30 minutes, one day consists of 48 time periods). The supplementation data range specification parameter is also written in a file stored in the storage unit 4, for example, and the power generation amount supplementation data acquisition unit 601 can read the supplementation data range specification parameter from this file.

[0021] The Complemented Power Generation Generation Unit 602 generates a compensated power generation amount based on the actual power generation values ​​up to the latest time period received from the Power Generation Complementary Data Acquisition Unit 601, weather data, and the power generation compensation model described later. Here, the compensated power generation amount is data that fills in the missing actual power generation values. The Complemented Power Generation Generation Unit 602 transmits the generated compensated power generation amount to the Power Generation Estimation Unit 604.

[0022] The weather data acquisition unit 603 acquires weather data from the weather data database. Based on the estimated time data range specification parameter, the weather data acquisition unit 603 extracts a predetermined range of weather data from the acquired data and transmits it to the power generation estimation unit 604. Here, the estimated time data range specification parameter is a parameter for specifying a predetermined range of data (for example, 10 days' worth) up to the estimated time period, and the estimated time period is the time period for which the power generation amount is to be estimated (for example, one day from a certain date and time onward). The estimated time data range specification parameter is also written in a file stored in the storage unit 4, for example, and the weather data acquisition unit 603 can read the estimated time data range specification parameter from this file.

[0023] The power generation estimation unit 604 estimates the amount of renewable energy power generated during the estimated time period based on the supplemented power generation amount received from the supplemented power generation amount generation unit 602, the weather data up to the estimated time period received from the weather data acquisition unit 603, and the power generation amount estimation model. The power generation estimation unit 604 transmits the estimated power generation amount data to the output unit 605.

[0024] The output unit 605 performs post-processing, such as clipping, on the data received from the power generation estimation unit 604, and outputs the processed data as power generation estimation data for the estimated time period. For example, by performing clipping processing, which converts the received power generation value to the rated output value if it exceeds the rated output value, or to 0 if it is negative, the output unit 605 limits the value of the power generation estimation data to a range from 0 to the rated output value. The output data is, for example, written to a file and stored in the storage unit 4.

[0025] The power generation estimation process performed by the processing unit 6 will be explained below with reference to Figures 3 to 8. Figure 8 is a flowchart showing an example of the procedure for power generation estimation.

[0026] First, the power generation data acquisition unit 601 acquires actual power generation values ​​from the power generation DB and weather data from the weather data DB, and extracts the actual power generation values ​​and weather data using the data range specification parameter for supplementation (step S10).

[0027] The actual power generation figures may be either data on the power generation of each wind turbine in the case of a wind power plant, or the total power generation of the entire power plant. Similarly, if the power plant is a solar power plant, the actual power generation figures may be either data on the power generation of each solar panel in the power plant, or the total power generation of the entire power plant.

[0028] Figure 3 shows an example of actual power generation data obtained from the power generation database. In this example, the power plant is a wind power plant with 10 wind turbines, and the actual power generation data measured by the power generation sensor of each wind turbine is shown for each time period. Here, the time periods are in 30-minute intervals. For example, the row in the "Date and Time" column, "2023-01-01 00:30:00," shows the actual power generation data [kWh] as the integral of the power [kW] from 00:00:00 to 00:29:59 on January 1, 2023. Note that the time intervals and "Date and Time" settings are not limited to these and can be anything.

[0029] As shown in Figure 3, the actual power generation figures may contain missing data indicated by "NA". Such missing data can occur, for example, if the SCADA (Supervisory Control And Data Acquisition) system that measures the actual power generation figures for each wind turbine fails to acquire the data correctly.

[0030] Figure 4 shows an example of meteorological data obtained from a meteorological data database. When estimating wind power generation, the items of meteorological data obtained include wind speed information for the mesh containing the location of the power plant to be estimated, and typically includes wind speed information for each pressure level. A mesh is a region in which the area targeted for observation of NWP and meteorological data is divided into squares ranging from a few kilometers to several hundred kilometers on each side.

[0031] When estimating wind power generation, the meteorological data items may include items such as wind direction and temperature in addition to those mentioned above. If the meteorological data items include data on the U component (east-west component) and V component (north-south component) of wind speed instead of wind speed magnitude and wind direction, the wind speed magnitude, etc., can be calculated from this data and added to the items.

[0032] When estimating solar power generation, the meteorological data items shall include items such as total horizontal solar radiation intensity, upper-level horizontal solar radiation intensity, sun direction, surface albedo (the percentage of sunlight reflected by the ground surface), temperature, and wind speed for the mesh containing the location of the power plant to be estimated. In this embodiment, the number of meteorological data items is set at 300, but the number of items is not limited to this.

[0033] Furthermore, while the example in Figure 4 assumes data generated by concatenating numerical weather forecast data at 30-minute intervals, estimating the next day's data, the granularity and number of estimation intervals, the timing of estimation (estimated execution time), the lead time of estimation (time difference from the estimated execution time), and the method of concatenating the data are not limited to this and may be any other. However, in this embodiment, it is assumed that the weather data obtained from the weather data DB does not contain missing values ​​(NA), such as actual power generation values. Unless otherwise specified, the following explanation will focus on estimating wind power generation.

[0034] Figure 5 shows an example of actual power generation data (Figure 5(a)) and weather data (Figure 5(b)) extracted using the supplementary data range specification parameter. In Figure 5, the most recent time period is the row where "Date and Time" is "2023-12-31 11:30:00". Also, in Figure 5, the supplementary data range specification parameter is set to 10 days.

[0035] Next, the interpolated power generation unit 602 generates interpolated power generation based on the actual power generation values ​​up to the latest time period, weather data, and the power generation interpolation model (step S11). For example, the power generation interpolation model is a machine learning-prepared linear regression model in which the cube of the wind speed at each pressure level in the weather data is the explanatory variable and the power generation amount of each wind turbine (or the total power generation amount of the power plant) is the dependent variable. This takes into account the physical relationship that the power generation amount of wind power is proportional to the cube of the wind speed.

[0036] The interpolated power generation unit 602 interpolates the data by replacing the missing values ​​in the actual power generation values ​​with the power generation values ​​output by the power generation interpolation model, thereby generating the interpolated power generation. At this time, if the value of the power generation output by the power generation interpolation model exceeds the rated output value, a clipping process is performed to convert it to the rated output value, and if it is negative, it is converted to 0. In other words, the interpolated power generation unit 602 limits the value of the power generation to a range from 0 to the rated output value.

[0037] The power generation interpolation model can be any model based on the physical relationship between meteorological data and power generation. For example, if power curve data representing the correspondence between wind speed [m / s] and power generation [kW] (instantaneous value of power generation), as shown in Figure 6, is available, this can be used as the power generation interpolation model.

[0038] Next, the weather data acquisition unit 603 acquires weather data from the weather data DB and extracts the weather data using the estimated time data range specification parameter (step S12). In this embodiment, the value of the estimated time data range specification parameter is set to the same value as the complementary data range specification parameter value (for example, 10 days).

[0039] Next, the power generation estimation unit 604 estimates the amount of renewable energy generated during the estimated time period based on the interpolated power generation up to the latest time period, weather data up to the estimated time period, and the power generation estimation model (step S13). The power generation estimation model is a neural network model based on LSTM (Long Short Term Memory), but other models (e.g., recurrent neural network, state-space model, transformer, etc.) may also be used.

[0040] Figure 7 shows an example of input data for a power generation estimation model. The input data consists of interpolated power generation data as shown in Figure 7(a) and meteorological data as shown in Figure 7(b). When the interpolation data range specification parameter and the estimation time data range specification parameter have the same value, the range of the interpolated power generation data (first range) and the range of the meteorological data (second range) have the same time series length, although the dates and times are different. Therefore, these data can be concatenated horizontally to form the input data. In the example in Figure 7, both the first and second ranges are 10 days (480 frames), and the second range is shifted 36.5 hours later than the first range. Also, the last day (48 frames) of the second range corresponds to the estimation time period.

[0041] In this embodiment, the number of hidden layers in the LSTM is set to, for example, approximately the number of features in the input data (300). The LSTM outputs data that is the same length as the time series length of the input data (480 frames) and has the same number of features as the number of hidden layers. Therefore, in an LSTM-based model, the last 48 frames (1 day) of the time series direction of the LSTM output data are extracted. The extracted data is input to a fully connected layer with an input layer having 300 nodes and an output layer having the same number of nodes as the wind turbines (10), and data is output that has a time series length equal to the length of the estimated time period and features corresponding to the power generation amount of each wind turbine. Note that the number of hidden layers in the LSTM may be other than 300.

[0042] Finally, the output unit 605 performs post-processing on the data output by the power generation estimation unit 604 and outputs it as estimated power generation data for the estimated time period (step S14). The post-processing involves clipping the data as described above. Also, for example, when estimating the total power generation of the entire power plant, the estimated values ​​for each wind turbine are added up and output as the total estimated power generation value.

[0043] Thus, according to the information processing device 1 of this embodiment, even when there are gaps in the actual values ​​of the amount of electricity generated by renewable energy sources, the amount of electricity generated can be estimated with high accuracy.

[0044] (Second Embodiment) Next, a second embodiment will be described.

[0045] This embodiment generates a power generation estimation model, as well as a complementary model of weather data and a complementary model of power generation data used in its generation. In the following description of the second embodiment, the description of parts that are the same as those of the first embodiment will be omitted, and the parts that differ from the first embodiment will be described.

[0046] Each power generation estimation model in this embodiment uses, along with (or in place of) the weather data described in the first embodiment, actual weather data measured by a weather sensor installed at a power plant with renewable energy sources. Since actual weather data, like actual power generation data, contains missing values, data interpolation is performed. In addition to missing values, actual power generation data and actual weather data in this embodiment are assumed to contain abnormal values ​​that are not normal, and data interpolation is also performed for these abnormal values.

[0047] In this embodiment, each model is generated by machine learning. Below, the training of the model that complements actual weather data is referred to as the first training, the training of the model that complements actual power generation data is referred to as the second training, and the training of the power generation estimation model is referred to as the third training. The period of data used to train each model is referred to as the training period. The training period may be the same for each model, or it may differ for each model.

[0048] Figure 9 shows an example of the functional configuration of the processing unit 6 according to this embodiment. The difference from the first embodiment is that the processing unit 6 further includes a first learning-time data acquisition unit 611, a normal period extraction unit for actual weather data 612, a weather data completion model learning unit 613, a weather data completion model memory reading unit 614, a second learning-time completed weather data generation unit 615, a second learning-time data acquisition unit 621, a normal period extraction unit for power generation 622, a power generation completion model learning unit 623, a power generation completion model memory reading unit 624, a third learning-time completed power generation generation unit 625, a third learning-time label generation unit 630, a third learning-time data acquisition unit 631, a power generation estimation model learning unit 632, a power generation estimation model memory reading unit 633, a weather data acquisition unit 641, a completed weather data generation unit 642, and an estimation-time label generation unit 650.

[0049] The first learning-time supplementation data acquisition unit 611 acquires weather performance values ​​for the learning period of the first learning from the weather performance value DB, acquires weather data for the learning period from the weather data DB, and transmits these to the weather performance value normal period extraction unit 612. Here, the weather data DB and the weather performance value DB are, for example, databases built on an external server device or the like of the information processing device 1. Furthermore, the weather data DB may be the same as the database in the first embodiment, or a different one.

[0050] The weather data stored in the weather data database consists of data measured by weather sensors installed at power plants with renewable energy sources, for example. These weather sensors include wind speed sensors, wind direction sensors, temperature sensors, etc.

[0051] The weather performance value normal period extraction unit 612 extracts weather performance values ​​for normal periods from the weather performance values ​​received from the first learning-time supplementation data acquisition unit 611. A normal period is a period in which the weather performance values ​​do not contain any missing or abnormal values. The normal period is also an example of the first period. The weather performance value normal period extraction unit 612 determines abnormal values ​​in the weather performance values ​​based on predetermined abnormal value determination conditions and obtains abnormal value determination information. The weather performance value normal period extraction unit 612 also determines any missing values ​​in the weather performance values ​​and obtains missing value determination information. Based on the abnormal value determination information and missing value determination information, the weather performance value normal period extraction unit 612 creates supplementation target information that indicates the areas in the weather performance values ​​that have abnormal or missing values ​​and extracts weather performance values ​​and weather data for normal periods that are not included in the supplementation target from the received data.

[0052] The weather performance value normal period extraction unit 612 transmits the weather performance values ​​and weather data for the normal period to the weather performance value completion model learning unit 613. It also transmits the weather performance values ​​and weather data for the learning period, as well as the information to be completed during the second learning period to the weather performance value generation unit 615, and transmits the information to be completed during the third learning period to the label generation unit 630. The weather performance value normal period extraction unit 612 may be configured to either determine only missing values ​​and extract periods without missing values ​​as normal periods, or to determine only abnormal values ​​and extract periods without abnormal values ​​as normal periods.

[0053] The weather performance data completion model learning unit 613 generates a weather performance data completion model using the weather data for normal periods received from the weather performance data normal period extraction unit 612 as input data for the first learning phase, and the weather performance data for normal periods as ground truth data for the first learning phase. The weather performance data completion model learning unit 613 also transmits the generated weather performance data completion model to the weather performance data completion model storage and reading unit 614.

[0054] The weather performance data completion model storage and reading unit 614 stores and reads the weather performance data completion model received from the weather performance data completion model learning unit 613. Various parameters of the weather performance data completion model are, for example, written in a file and stored in the storage unit 4. The weather performance data completion model storage and reading unit 614 also transmits the read weather performance data completion model to the second learning-completed weather performance data generation unit 615 and the completed weather performance data generation unit 642.

[0055] The second learning-time supplemented weather performance value generation unit 615 generates supplemented weather performance values ​​based on the weather performance values, weather data, information to be supplemented, and the weather performance value supplementation model read by the weather performance value supplementation model storage and reading unit 614 during the learning period of the second learning. Specifically, the second learning-time supplemented weather performance value generation unit 615 extracts the weather performance values ​​to be supplemented from the information to be supplemented, inputs these into the weather performance value supplementation model, and calculates the weather performance values ​​for supplementation. This generates supplemented weather performance values ​​by supplementing the weather performance values ​​to be supplemented. The second learning-time supplemented weather performance value generation unit 615 also transmits the generated supplemented weather performance values ​​to the second learning-time supplementation data acquisition unit 621.

[0056] The second learning-time supplementation data acquisition unit 621 acquires the actual power generation values ​​for the learning period of the second learning from the power generation DB, and the supplemented weather actual values ​​and weather data for the learning period of the second learning from the second learning-time supplemented weather actual value generation unit 615. The second learning-time supplementation data acquisition unit 621 also transmits the actual power generation values, supplemented weather actual values, and weather data to the power generation normal period extraction unit 622. Here, the power generation DB is, for example, a database built on an external server device or the like of the information processing device 1. The power generation DB may be the same as the database in the first embodiment, or a different one.

[0057] The normal power generation period extraction unit 622 receives actual power generation values, supplemented weather actual values, and weather data from the second learning-time supplementation data acquisition unit 621, determines abnormal values ​​in the actual power generation values ​​based on predetermined abnormal value determination conditions, and obtains abnormal value determination information. The normal power generation period extraction unit 622 also determines any missing values ​​in the actual power generation values ​​and obtains missing value determination information. Based on the abnormal value determination information and missing value determination information, the normal power generation period extraction unit 622 creates power generation supplementation target information that indicates supplementation targets where abnormal values ​​or missing values ​​exist in the actual power generation values, and extracts actual power generation values, supplemented weather actual values, and weather data for normal periods that are not supplementation targets from the received data based on this information.

[0058] The normal power generation period extraction unit 622 transmits the supplemented weather performance values ​​and weather data for the normal period to the power generation amount supplementation model learning unit 623, and transmits the power generation performance values, supplemented weather performance values, weather data, and power generation amount supplementation target information to the third learning supplemented power generation amount generation unit 625. The normal power generation period extraction unit 622 may be configured to determine only missing values ​​and extract periods without missing values ​​as normal periods, or to determine only abnormal values ​​and extract periods without abnormal values ​​as normal periods.

[0059] The power generation amount interpolation model learning unit 623 generates a power generation amount interpolation model using either or both of the interpolated weather performance values ​​and weather data for the normal period received from the power generation amount normal period extraction unit 622 as input data for the second learning, and the power generation amount performance values ​​for the normal period as ground truth data for the second learning. The power generation amount interpolation model learning unit 623 also transmits the generated power generation amount interpolation model to the power generation amount interpolation model storage and reading unit 624.

[0060] The power generation amount interpolation model storage and reading unit 624 stores and reads the power generation amount interpolation model received from the power generation amount interpolation model learning unit 623. Various parameters of the power generation amount interpolation model are, for example, written in a file and stored in the storage unit 4. The power generation amount interpolation model storage and reading unit 624 also transmits the read power generation amount interpolation model to the third learning-interpolated power generation amount generation unit 625 and the interpolated power generation amount generation unit 602.

[0061] The third learning-time supplemented power generation amount generation unit 625 supplements the actual power generation amount to be supplemented during the learning period and generates the supplemented power generation amount for the learning period of the third learning. Specifically, it receives the actual power generation amount for the learning period, supplemented weather actual values, weather data, and power generation amount supplement target information from the power generation amount normal period extraction unit 622, and generates the supplemented power generation amount based on the power generation amount supplementation model read by the power generation amount supplementation model storage read unit 624. The third learning-time supplemented power generation amount generation unit 625 also transmits the generated supplemented power generation amount to the third learning-time data acquisition unit 631.

[0062] The third learning label generation unit 630 generates supplemental label data for the learning period. Specifically, the third learning label generation unit 630 generates supplemental label data based on the supplemental information for the learning period received from the weather performance normal period extraction unit 612 and the power generation amount supplemental information for the learning period received from the power generation amount normal period extraction unit 622. The third learning label generation unit 630 also transmits the generated supplemental label data to the third learning data acquisition unit 631.

[0063] The third learning data acquisition unit 631 acquires weather data for the learning period, supplemented weather actual values, and supplemented power generation amount from the third learning supplemented power generation amount generation unit 625, and acquires supplement target label data for the learning period from the third learning label generation unit 630. Based on the third learning data range specification parameter, the third learning data acquisition unit 631 extracts data from the acquired learning period multiple times, as described later, data from the first range and data from the second range, and creates multiple input data sets for the third learning. The third learning data acquisition unit 631 also creates multiple ground truth data sets from multiple second range weather data sets. During the learning period, the third learning data acquisition unit 631 creates multiple datasets of input data and ground truth data and transmits them to the power generation estimation model learning unit 632.

[0064] The power generation estimation model learning unit 632 generates a power generation estimation model using the input data and ground truth data dataset received from the third learning data acquisition unit 631. The power generation estimation model learning unit 632 then transmits the generated model to the power generation estimation model storage and retrieval unit 633.

[0065] The power generation estimation model storage and retrieval unit 633 stores and retrieves the power generation estimation model received from the power generation estimation model learning unit 632. Various parameters of the power generation estimation model are, for example, written in a file and stored in the storage unit 4. The power generation estimation model storage and retrieval unit 633 also transmits the retrieved power generation estimation model to the power generation estimation unit 604.

[0066] The weather data acquisition unit 641 for supplementing weather performance data acquires weather data from the weather data DB and weather performance data from the weather performance data DB, and extracts data within a predetermined range from the acquired data based on the supplement data range specification parameter. Here, the weather data DB and weather performance data DB are, for example, databases built on an external server device or the like of the information processing device 1. Furthermore, these databases may be the same as or different from the database from which the first learning-time supplement data acquisition unit 611 acquires data.

[0067] The weather data acquisition unit 641 for supplementing actual weather data determines any missing values ​​in the extracted actual weather data and obtains missing value determination information. The weather data acquisition unit 641 also transmits the actual weather data, weather data, and missing value determination information up to the latest time period to the supplemented actual weather data generation unit 642, and transmits the missing value determination information up to the latest time period to the estimated time label generation unit 650.

[0068] The Complemented Weather Data Generation Unit 642 generates compensated weather data up to the latest time period. Specifically, the Complemented Weather Data Generation Unit 642 generates compensated weather data by supplementing the data to be supplemented based on the weather data, weather data, missing period determination information, and the weather data supplementation model read by the weather data supplementation model storage and reading unit 614. The Complemented Weather Data Generation Unit 642 also transmits the generated compensated weather data and weather data to the Power Generation Data Acquisition Unit 601.

[0069] The power generation amount supplementation data acquisition unit 601 acquires data for generating supplemented power generation amounts. The difference from the first embodiment is that it further acquires supplemented weather performance values ​​up to the latest time period, and determines any missing values ​​in the extracted power generation performance values ​​and obtains missing value determination information indicating the determined missing values. Furthermore, the power generation amount supplementation data acquisition unit 601 transmits the weather performance values ​​up to the latest time period and the missing value determination information to the supplemented weather performance value generation unit 642, and further transmits the missing value determination information up to the latest time period to the estimated time label generation unit 650.

[0070] The supplemented power generation unit 602 generates the supplemented power generation amount. The difference from the first embodiment is that it further transmits the supplemented weather data up to the latest time period to the power generation estimation unit 604.

[0071] The estimated time label generation unit 650 generates missing label data up to the latest time period. Specifically, it generates missing label data up to the latest time period based on the missing data judgment information received from the weather performance data acquisition unit 641 and the missing data judgment information received from the power generation amount acquisition unit 601. The estimated time label generation unit 650 transmits the generated missing label data to the power generation amount estimation unit 604.

[0072] The power generation estimation unit 604 estimates the amount of power generated by renewable energy sources during the estimated time period. The difference from the first embodiment is that it further receives the supplemented weather data up to the latest time period generated by the supplemented power generation generation unit 602 and the missing label data generated by the estimation time label generation unit 650, and uses these to estimate the amount of power generated.

[0073] The learning process for each model and the power generation estimation process performed by the processing unit 6 will be explained below with reference to Figures 10 to 16.

[0074] Figure 15 is a flowchart showing an example of the training process for each model. First, the first training data acquisition unit 611 acquires weather performance data for the training period from the weather performance data DB and weather data for the training period from the weather data DB (step S20). In the following explanation, the training period will be two years, from the 0:30 frame on January 1, 2021 to the 0:00 frame on January 1, 2023.

[0075] The acquired weather data are, for example, features other than the power generation amount of the SCADA system that measures data for each wind turbine, such as actual values ​​of wind speed and wind direction. In this embodiment, an example of a power plant with 10 wind turbines will be described, similar to the first embodiment. In the following, actual wind speed values ​​will be used as weather data, but other features may be used. Figure 10(a) shows an example of weather data for the learning period. As shown in Figure 10(a), the data includes missing values ​​(NA) and consecutive zeros. On the other hand, the weather data in the weather data DB has 300 items, similar to the one described in Figure 4, and does not contain missing values.

[0076] Next, the weather data normal period extraction unit 612 determines whether there are missing or abnormal weather data (step S21). For example, the weather data normal period extraction unit 612 determines that an abnormal value exists if the wind speed data is the same for three consecutive frames. However, the criteria for determining abnormal values ​​are not limited to this and may be any.

[0077] Figure 11(a) shows an example of anomaly detection information corresponding to the data in Figure 10(a). As shown in Figure 11(a), the anomaly detection information is table data that has a value of "True" when it is determined to be abnormal and "False" when it is determined to be normal.

[0078] The weather data normal period extraction unit 612 further determines missing weather data (step S21). Figure 11(b) shows an example of missing data determination information corresponding to the data in Figure 10(a). As shown in Figure 11(b), the missing data determination information is table data with a value of "True" if it is determined to be missing, and "False" if it is determined to be normal.

[0079] The weather performance value normal period extraction unit 612 creates supplementary information indicating that the weather performance value is an abnormal value or missing by taking the union of abnormal value judgment information and missing value judgment information (step S22). Figure 12 shows an example of supplementary information created from the data tables in Figures 11(a) and 11(b). As shown in Figure 12, the supplementary information is table data that has the value "True" if it is a target for supplementation and "False" if it is not a target for supplementation.

[0080] The weather performance value normal period extraction unit 612 extracts weather performance values ​​and weather data for the normal period (the part corresponding to False in the supplementation target information) for each wind turbine from the learning period based on the information to be supplemented (step S23).

[0081] Next, the weather performance data interpolation model learning unit 613 generates a weather performance data interpolation model using machine learning, with weather data from the normal period as input data and weather performance data from the normal period as ground truth data (step S24). For example, the weather performance data interpolation model learning unit 613 creates a linear regression model for each wind turbine, with wind speeds from pressure levels 1 to 10 (a total of 10 variables) from the weather data as explanatory variables and wind speed performance values ​​measured by the wind speed sensor as the dependent variable. Since the wind speed performance values ​​are thought to have a physical relationship with the wind speed in the weather data, it is possible to generate such a model.

[0082] Next, the weather performance value completion model storage and reading unit 614 stores the generated weather performance value completion model (step S25). In this embodiment, completion models for 10 wind speed performance values ​​are stored.

[0083] Next, the second learning-interpolated weather performance value generation unit 615 generates interpolated weather performance values ​​using the weather performance value interpolation model (step S26). In this embodiment, for each time frame to be interpolated, the wind speed at each pressure level in the weather data (10 variables in total) is input to each learned interpolation model, and the wind speed of each corresponding wind turbine is output. As a post-processing step, if the output wind speed is a negative value, a clipping process may be performed to convert it to 0. The wind speed gaps for each wind turbine to be interpolated are interpolated using the post-processed data, and interpolated wind speed performance values ​​are generated. Figure 10(b) shows an example of interpolated wind speed performance values ​​corresponding to the wind speed performance values ​​in Figure 10(a).

[0084] Next, the second learning-time interpolation data acquisition unit 621 acquires the actual power generation values ​​for the learning period from the power generation DB, in addition to the weather data and interpolated actual weather values ​​for the learning period (step S27). Figure 13 shows an example of the actual power generation values ​​for the learning period. As shown in Figure 13, the data of the actual power generation values ​​includes missing values ​​(NA) and places where consecutive zeros appear abnormal, but these places are not necessarily the same as the places where there are missing values ​​or abnormal values ​​in the actual weather values.

[0085] Next, the normal power generation period extraction unit 622 determines whether there are missing or abnormal values ​​in the actual power generation values ​​(step S28). For example, the normal power generation period extraction unit 622 determines that an abnormal value is present if the actual wind speed value is the same for three consecutive time slots. However, the conditions for determining abnormal values ​​are not limited to this and can be anything. For example, for weather data in a certain time slot, if the majority of the pressure surface is estimated to have a wind speed of or greater than the cut-in wind speed, and the actual wind speed value corresponding to that time slot is also greater than or equal to the cut-in wind speed, then it may be determined that an abnormal value is present if the actual power generation value corresponding to that time slot is 0. The cut-in wind speed is the wind speed at which the wind turbine reaches a state where it can start generating power after it has started rotating.

[0086] The normal power generation period extraction unit 622 further determines if there are any missing values ​​in the actual power generation data (step S28). Then, by taking the union of abnormal value determination information indicating abnormal values ​​in power generation and missing value determination information indicating missing values, it creates supplementary information indicating that the actual power generation data is an abnormal value or missing (step S29). Based on the supplementary information, the normal power generation period extraction unit 622 extracts the actual power generation data and weather data for the normal period for each wind turbine from the learning period (step S30).

[0087] Next, the power generation interpolation model learning unit 623 uses either or both of the weather data and / or actual weather data from the normal period as input data, and the actual power generation values ​​from the normal period as ground truth data to generate a power generation interpolation model by machine learning (step S31). For example, the power generation interpolation model learning unit 623 creates a linear regression model for each wind turbine, with the cube values ​​of either or both of the wind speeds at each pressure level in the weather data and the corresponding interpolated actual wind speed values ​​as explanatory variables, and power generation as the dependent variable. Alternatively, by fitting the data to a function such as a logistic curve, an interpolation model may be generated with the wind speed at the pressure level with the highest correlation between actual wind speed values ​​and wind speed, or the interpolated actual wind speed values, as explanatory variables, and the power generation of the wind turbine as the dependent variable.

[0088] Next, the power generation amount interpolation model storage and reading unit 624 stores the generated power generation amount interpolation model (step S32). In this embodiment, 10 power generation amount interpolation models are stored.

[0089] Next, the third-stage learned-upgraded power generation unit 625 generates an upgraded power generation amount using the power generation amount interpolation model (step S33). In this embodiment, for each time frame to be interpolated, the explanatory variables used to generate the power generation amount interpolation model are input to each trained interpolation model, and the corresponding power generation value for each wind turbine is output. As a post-processing step, if the output wind speed is equal to or greater than the rated output value, a clipping process may be performed to convert it to the rated output value, or if it is a negative value, it may be converted to 0. The data after post-processing is used to interpolate the power generation gaps for each wind turbine to be interpolated, and the upgraded power generation amount is generated.

[0090] On the other hand, the third learning label generation unit 630 creates supplemental label data based on the supplemental information for the actual power generation values ​​and the supplemental information for the actual weather values ​​during the learning period (step S34). The supplemental label data is, for example, for each time frame of each wind turbine, label data that is 1 if the actual power generation value is subject to supplementation and 0 otherwise, and label data that is 1 if the actual wind speed value is subject to supplementation and 0 otherwise. Figure 14 shows an example of the supplemental label data created by the third learning label generation unit 630.

[0091] Furthermore, the label data to be supplemented may be set to 2 if both the actual power generation value and the actual wind speed value are to be supplemented, to 1 if one of them is to be supplemented, and to 0 if neither is to be supplemented. Alternatively, instead of creating label data for each wind turbine, the label may be set to 1 if a predetermined number of wind turbines or more are to be supplemented in each time frame for the entire power plant, and to 0 if they are not to be supplemented. In addition, label data for missing values ​​and outliers may be created separately, or label data for only missing values ​​or outliers may be created, etc.

[0092] Next, the third learning data acquisition unit 631 acquires data to be used for the third learning based on the third learning data range specification parameter (step S35). The data used for the third learning includes multiple datasets of input data and corresponding ground truth data during the learning period.

[0093] The input data consists of the first range of interpolated power generation, the first range of interpolated weather data, the second range of interpolated label data, and the second range of weather data. In this embodiment, the lengths of the first and second ranges are the same (e.g., 10 days), and the second range is shifted 36.5 hours later than the first range. Although the dates and times of the data in the first and second ranges are different, their time series lengths are equal, so these data can be concatenated to form the input data.

[0094] The correct data corresponding to the input data above is the actual power generation value for the latter half of the second range, which does not overlap with the first range. For example, it is the actual power generation value for the last day of the second range.

[0095] The third learning data acquisition unit 631 creates multiple datasets of input data and ground truth data for each of the multiple first ranges during the learning period. If the length of the first and second ranges is 10 days, and the first range is the 10 days from 12:00 on each day of the learning period, and the range of the ground truth data corresponding to this is the last day of the second range, then the first range of input data and the range of the ground truth data corresponding to it will be as follows.

[0096] Past input data: January 1, 2021, 12:00~ January 11, 2021 11:30 Corresponding correct answer data: January 12, 2021, 00:30~ January 13, 2021 00:00 Second oldest input data: January 2, 2021, 12:00~ January 12, 2021 11:30 Corresponding correct answer data: January 13, 2021, 00:30~ January 14, 2021 00:00 ... Most recent input data: December 21, 2022, 12:00~ December 31, 2022 11:30 Corresponding correct answer data: January 1, 2023, 00:30~ January 2, 2023 00:00

[0097] In this case, 365 × 2 - 11 = 719 sets of input data and ground truth data are created. The number of feature items in the input data is 340 (the sum of actual values ​​of power generation and wind speed for 10 wind turbines, complementary label data for power generation and wind speed, and 300 items of weather data). The number of feature items in the ground truth data is 10 (actual values ​​of power generation for 10 wind turbines). The third training data acquisition unit 631 normalizes and standardizes this data for each feature, then concatenates it to create a 3D tensor of (number of datasets, number of time frames per set, number of feature items). In the above example, the input data dataset is a 3D tensor of (719, 480, 340), and the ground truth data dataset is a 3D tensor of (719, 48, 10).

[0098] Next, the power generation estimation model learning unit 632 generates a power generation estimation model using machine learning with the input data and the ground truth data dataset (step S36). In this embodiment, mini-batch learning is performed by dividing the 3D tensor described above into mini-batches as described later, but batch learning may also be performed using the 3D tensor without dividing it. Furthermore, the power generation estimation model may be a two-stage model consisting of a first-stage model that estimates the actual wind speed value for each wind turbine from, for example, a source other than the actual power generation value, and a second-stage model that estimates the power generation amount for each wind turbine from the wind speed estimated by the first-stage model. Also, the power generation estimation model may be configured for each wind turbine.

[0099] The LSTM used in the power generation estimation model outputs data with a length equal to the time series length of the input data and a number of features equal to the number of hidden layers. For example, if the number of hidden layers in the LSTM is approximately the number of features in the input data (300) and the mini-batch size is 70, the LSTM converts the input data, a 3D tensor of (70,480,340), into a 3D tensor of (70,480,300). Therefore, in an LSTM-based model, the last 48 frames (one day's worth) of the time series direction of the LSTM output data are extracted from the converted 3D tensor to obtain a 3D tensor of (70,48,300). The obtained data is input to a fully connected layer with 300 nodes in the input layer and 10 nodes in the output layer, and a 3D tensor of (70,48,10) is output, with a time series length equal to the length of the estimated time period and features corresponding to the power generation of each wind turbine. In this embodiment, the number of hidden layers in the LSTM was set to approximately the number of features in the input data (300), but the number of hidden layers is not limited to this.

[0100] Next, the loss is calculated from the output data and the ground truth data (if the mini-batch size is 70, both are 3D tensor data of (70,48,10)). The gradient is then calculated using backpropagation to determine how the LSTM weights and bias parameters change. These parameters are then updated and optimized using stochastic gradient descent.

[0101] In this embodiment, Adam (Adaptive moment) is used as the optimization method and L1 loss due to absolute error is used as the loss function, but other optimization methods and loss functions may be used. For example, the variation in the amount of power generated by a wind turbine with respect to wind speed is small below the cut-in wind speed and above the rated wind speed (the wind speed that generates the rated output of the wind turbine), in accordance with the power curve, and becomes large between the cut-in wind speed and the rated wind speed. Therefore, if the power generation estimation model is the two-stage model described above, the accuracy of estimating the actual wind speed may be improved by amplifying the loss in the first-stage model between the cut-in wind speed and the rated wind speed. For this purpose, for example, a function F of the power curve such as a logistic curve may be used, and the error between the actual wind speed yn (normalized or standardized value) and its estimated value yp may be calculated using the L1 loss for F(yn) and F(yp).

[0102] Next, the power generation estimation model storage and reading unit 633 stores the generated power generation estimation model (step S37).

[0103] Figure 16 is a flowchart showing an example of the procedure for estimating the amount of power generated. The differences from Figure 8 will be explained below. Note that the operations in steps S43, S45, and S47 are the same as in steps S11, S12, and S14 in Figure 8.

[0104] The weather data acquisition unit 641 for supplementing actual weather data extracts actual weather data and weather data using the supplement data range specification parameter (step S40).

[0105] The supplemented weather data generation unit 642 generates supplemented weather data up to the latest time period (step S41).

[0106] The power generation data acquisition unit 601 extracts actual power generation values ​​using the data range specification parameter for supplementation (step S42).

[0107] The estimated time label generation unit 650 generates missing period label data up to the latest time period (step S44).

[0108] The power generation estimation unit 604 estimates the amount of power generated during the estimated time period (step S46). The difference from Figure 8 is that the power generation is further estimated using the interpolated weather data and missing label data.

[0109] Thus, according to the information processing device 1 of this embodiment, by generating a power generation estimation model, and generating complementary models for weather data and power generation data used in its generation, it is possible to accurately estimate power generation even when there are abnormal values ​​or omissions in the weather data and power generation data of renewable energy sources.

[0110] Furthermore, the programs executed in each functional unit of each embodiment may be configured to be provided as installable or executable files recorded on a computer-readable recording medium such as a CD-ROM, flexible disk (FD), CD-R, or DVD (Digital Versatile Disk).

[0111] The program may be configured to be stored on a computer connected to a network such as the Internet and provided by downloading it via the network, or it may be configured to be provided or distributed via a network such as the Internet. Alternatively, the program may be configured to be pre-installed in ROM or the like.

[0112] The program executed in each device of each embodiment has a modular configuration that includes the functional units described above. In actual hardware, the CPU (processor) reads the program from the storage medium and executes it, loading the units onto the main memory, and creating each functional unit on the main memory.

[0113] Each function of the embodiments described above can be realized by one or more processing circuits. Hereinafter, "processing circuit" as used herein includes processors programmed to execute each function by software, such as processors implemented by electronic circuits, as well as devices such as ASICs (Application Specific Integrated Circuits), DSPs (digital signal processors), FPGAs (field programmable gate arrays), and conventional circuit modules designed to execute each function described above.

[0114] Although various embodiments of the present invention have been described above, these embodiments are presented as examples only and are not intended to limit the scope of the invention. These novel embodiments can be implemented in a variety of other forms, and various combinations, omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These novel embodiments and their variations are included in the scope and spirit of the invention, as well as in the claims of the invention and its equivalents. Furthermore, components from different embodiments and modifications may be combined as appropriate.

[0115] (Note) An example of the configuration of the embodiment is described below.

[0116] (Configuration Example 1) We obtain weather data and actual power generation values ​​measured by power generation sensors. Based on the aforementioned weather data and the aforementioned actual power generation values, a supplemented power generation amount is generated by filling in the missing values ​​in the aforementioned actual power generation values. Based on the aforementioned supplemented power generation amount, the weather data, and the power generation estimation model, the amount of power generated by renewable energy sources is estimated. A processing unit is provided, Information processing device. (Configuration example 2) The processing unit performs training on the power generation estimation model based on the weather data and the supplemented power generation amount during the training period. The information processing device described in Configuration Example 1. (Configuration Example 3) The aforementioned processing unit, From the aforementioned actual power generation values, extract the first period that does not include the aforementioned deficit, Based on the weather data for the first period and the actual power generation values, a power generation interpolation model is trained to generate the interpolated power generation amount. An information processing device as described in Configuration Example 1 or 2. (Configuration example 4) The aforementioned renewable energy source is a wind power generation device. The aforementioned power generation interpolation model is a linear regression model in which the cube of the wind speed data included in the weather data for the first period is used as the explanatory variable, and the actual power generation value for the first period is used as the dependent variable. The information processing device described in Configuration Example 3. (Configuration example 5) The explanatory variables include either or both the cube of wind speed data obtained from numerical weather forecasting and the cube of actual wind speed values ​​measured by wind speed sensors. The information processing device described in Configuration Example 4. (Configuration example 6) The processing unit restricts the value of the power generation output by the learned power generation interpolation model to a range from 0 to the rated output value. An information processing device as described in any one of configuration examples 3 to 5. (Configuration example 7) The processing unit extracts the first period based on the missing and abnormal values ​​of the actual power generation data. An information processing device as described in any one of configuration examples 3 to 6. (Configuration example 8) The processing unit generates the supplemented power generation amount based on a power curve that shows the correspondence between weather data and power generation amount. An information processing device as described in any one of Configuration Examples 1 to 7. (Configuration example 9) The aforementioned weather data is weather data obtained from numerical weather forecasting. An information processing device as described in any one of Configuration Examples 1 to 8. (Configuration example 10) The processing unit limits the value of the power generation output by the power generation estimation model to a range from 0 to the rated output value. An information processing device as described in any one of configuration examples 3 to 7. (Configuration Example 11) The aforementioned processing unit, From the aforementioned actual power generation values, extract the first period that does not include the aforementioned deficit, The weather data for the first period and the actual power generation values ​​are fitted to a predetermined function to generate a power generation interpolation model for generating the interpolated power generation amount. An information processing device as described in Configuration Example 1 or 2. (Configuration Example 12) The aforementioned processing unit, Extract the first period from the weather data measured by weather sensors, excluding any missing data. Based on the weather data and actual weather values ​​for the first period, a weather data interpolation model is trained to generate interpolated weather values ​​that fill in any missing information in the actual weather values. An information processing device as described in any one of Configuration Examples 1 to 11. (Configuration Example 13) The processing unit extracts the first period based on the missing and outlier values ​​of the actual weather data. The information processing device described in Configuration Example 12. (Configuration Example 14) The supplemented weather data is generated based on the weather data up to the latest time period and the actual weather data. An information processing device as described in Configuration Example 12 or 13. (Configuration example 15) An information processing method executed by a processing unit in an information processing device, The steps include acquiring weather data and actual power generation values ​​measured by power generation sensors, The steps include generating a supplemented power generation amount by filling in the gaps in the actual power generation amount based on the weather data and the actual power generation amount, The steps include: estimating the amount of electricity generated by renewable energy sources based on the supplemented power generation amount, the weather data, and the power generation estimation model; Information processing methods including (Configuration Example 16) Computers, We obtain weather data and actual power generation values ​​measured by power generation sensors. Based on the aforementioned weather data and the aforementioned actual power generation values, a supplemented power generation amount is generated by filling in the missing values ​​in the aforementioned actual power generation values. A processing unit that estimates the amount of electricity generated by renewable energy sources based on the supplemented power generation amount, the weather data, and the power generation estimation model. A program designed to function as such. [Explanation of Symbols]

[0117] 1. Information Processing Device 2 Input section 3 Display section 4 Storage section 5 Communications Department 6 Processing Unit 601 Data acquisition unit for supplementing power generation amount 602 Supplemented Power Generation Unit 603 Weather Data Acquisition Unit 604 Power generation estimation unit 605 Output section

Claims

1. We obtain weather data and actual power generation values ​​measured by power generation sensors. Based on the aforementioned weather data and the aforementioned actual power generation values, a supplemented power generation amount is generated by filling in the missing values ​​in the aforementioned actual power generation values. Based on the aforementioned supplemented power generation amount, the weather data, and the power generation estimation model, the amount of power generated by renewable energy sources is estimated. A processing unit is provided, Information processing device.

2. The processing unit performs training on the power generation estimation model based on the weather data and the supplemented power generation amount during the training period. The information processing apparatus according to claim 1.

3. The aforementioned processing unit, From the aforementioned actual power generation values, extract the first period that does not include the aforementioned deficit, Based on the weather data for the first period and the actual power generation values, a power generation interpolation model is trained to generate the interpolated power generation amount. The information processing apparatus according to claim 1.

4. The aforementioned renewable energy source is a wind power generation device. The power generation interpolation model is a linear regression model in which the cube of the wind speed data included in the weather data for the first period is used as the explanatory variable, and the actual power generation value for the first period is used as the dependent variable. The information processing apparatus according to claim 3.

5. The explanatory variables include either or both the cube of wind speed data obtained from numerical weather forecasting and the cube of actual wind speed values ​​measured by wind speed sensors. The information processing apparatus according to claim 4.

6. The processing unit restricts the value of the power generation output by the learned power generation interpolation model to a range from 0 to the rated output value. The information processing apparatus according to claim 3.

7. The processing unit extracts the first period based on the missing and abnormal values ​​of the actual power generation data. The information processing apparatus according to claim 3.

8. The processing unit generates the supplemented power generation amount based on a power curve that shows the correspondence between weather data and power generation amount. The information processing apparatus according to claim 1.

9. The aforementioned weather data is weather data obtained from numerical weather forecasting. The information processing apparatus according to claim 1.

10. The processing unit limits the value of the power generation output by the power generation estimation model to a range from 0 to the rated output value. The information processing apparatus according to claim 3.

11. The aforementioned processing unit, From the aforementioned actual power generation values, extract the first period that does not include the aforementioned deficit, The weather data for the first period and the actual power generation values ​​are fitted to a predetermined function to generate a power generation interpolation model for generating the interpolated power generation amount. The information processing apparatus according to claim 1.

12. The aforementioned processing unit, Extract the first period, which does not contain missing data, from the actual weather data measured by weather sensors. Based on the weather data and actual weather values ​​for the first period, a weather data interpolation model is trained to generate interpolated weather values ​​that fill in any missing information in the actual weather values. The information processing apparatus according to claim 1.

13. The processing unit extracts the first period based on the missing and abnormal values ​​of the actual weather data. The information processing apparatus according to claim 12.

14. The supplemented weather data is generated based on the weather data up to the latest time period and the actual weather data. The information processing apparatus according to claim 12.

15. An information processing method executed by a processing unit in an information processing device, The steps include acquiring weather data and actual power generation values ​​measured by power generation sensors, The steps include generating a supplemented power generation amount by filling in the gaps in the actual power generation amount based on the weather data and the actual power generation amount, The steps include: estimating the amount of electricity generated by renewable energy sources based on the supplemented power generation amount, the weather data, and the power generation estimation model; Information processing methods including

16. Computers, We obtain weather data and actual power generation values ​​measured by power generation sensors. Based on the aforementioned weather data and the aforementioned actual power generation values, a supplemented power generation amount is generated by filling in the missing values ​​in the aforementioned actual power generation values. A processing unit that estimates the amount of electricity generated by renewable energy sources based on the supplemented power generation amount, the weather data, and the power generation estimation model. A program designed to function as such.

Images