Photovoltaic power generation performance evaluation device, photovoltaic power generation performance evaluation method, and photovoltaic power generation performance

Abstract

To provide a photovoltaic power generation performance evaluation device, a photovoltaic power generation performance evaluation method, and a photovoltaic power generation performance evaluation program for reducing the preparation load of a learning model by a user, and for achieving highly accurate photovoltaic power generation performance evaluation.SOLUTION: According to one embodiment, a photovoltaic power generation performance evaluation device includes an explanatory variable preprocessing unit for preprocessing first weather data. The photovoltaic power generation performance evaluation device further includes a learning model generation unit that generates a learning model by machine learning using the first weather data as an explanatory variable and a result of simulating system power generation performance in the photovoltaic power plant as an objective variable. The photovoltaic power generation performance evaluation device further includes an acquisition part for acquiring second weather data and power generation amount data. Further, the photovoltaic power generation performance evaluation device includes a data processing unit that compares the system power generation performance with the actual power generation performance to evaluate the power generation performance of the photovoltaic power plant.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] An embodiment of the present invention relates to a photovoltaic power generation performance evaluation device, a photovoltaic power generation performance evaluation method, and a photovoltaic power generation performance evaluation program. [Background technology]

[0002] The power generation performance of a solar power generation system is easily affected by environmental factors such as weather conditions. Therefore, it is important to check on a daily basis whether the expected power generation performance is being achieved. Generally, power generation performance is evaluated using meteorological data such as sunlight and temperature at the time of power generation and actual power generation data. In addition, to evaluate whether power generation performance is declining, an evaluation method may be used that compares power generation data at the time of evaluation with power generation data calculated from past weather data. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent No. 6093465 Summary of the Invention [Problem to be solved by the invention]

[0004] In the above-mentioned system, the performance of the photovoltaic power generation equipment is evaluated using past power generation data. However, when there is no past power generation data or there is not enough data, such as immediately after the start of operation of the photovoltaic power plant, it is difficult to detect a decline in power generation performance using the above-mentioned evaluation method.

[0005] Furthermore, the above-mentioned system is based on the premise that data from periods when optimal power generation was performed is used as past power generation data, so if an evaluation is performed using power generation data from periods when optimal power generation was not performed, there is a possibility that an accurate evaluation will not be made.

[0006] In addition, a method has been proposed for evaluating the power generation performance of a solar power plant, even if no past power generation data exists, by using a trained model generated by machine learning with weather data as an explanatory variable and the results of simulating the power generation performance of the solar power plant as the target variable. However, creating a trained model requires a large amount of training data, which places a burden on the user. In addition, it is possible that the user does not have weather data for the solar power plant to be evaluated, and there is no data to input as an explanatory variable.

[0007] Therefore, this embodiment of the present invention proposes a photovoltaic power generation performance evaluation device, a photovoltaic power generation performance evaluation method, and a photovoltaic power generation performance evaluation program that reduce the burden on the user of creating a learning model and enable more accurate photovoltaic power generation performance evaluation. [Means for solving the problem]

[0008] According to one embodiment, the photovoltaic power generation performance evaluation device includes an explanatory variable preprocessing unit that preprocesses first weather data. Furthermore, the photovoltaic power generation performance evaluation device includes a learning model generation unit that generates a learning model by machine learning using the first weather data as an explanatory variable and a result of simulating the system power generation performance in the photovoltaic power plant as a target variable. Furthermore, the photovoltaic power generation performance evaluation device includes an acquisition unit that acquires second weather data and power generation amount data. Furthermore, the photovoltaic power generation performance evaluation device includes a data processing unit that compares the system power generation performance with the actual power generation performance and evaluates the power generation performance of the photovoltaic power plant. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is a schematic configuration diagram of a solar power plant 100 according to a first embodiment. [Figure 2] 1 is a block diagram showing the configuration of a solar power plant 100 according to a first embodiment. [Figure 3] 1 is a block diagram of a photovoltaic power generation performance evaluation device 200 in a first embodiment. [Figure 4] 3 is an example of weather data in the first embodiment. [Figure 5] 4 is an example of interpolation of meteorological data in the first embodiment. [Figure 6] 10 is an example in which the range of values ​​that can be taken by the weather data in the first embodiment is expanded. [Figure 7] 10 is an example of excluding weather data in the first embodiment. [Figure 8] 10 is an example of a flowchart for generating a learning model in the first embodiment. [Figure 9] 4 is an example of a flowchart for evaluating power generation performance in the first embodiment. [Figure 10] 10 is an example of a flowchart when updating a learning model in the second embodiment. [Figure 11] FIG. 2 is a hardware configuration diagram of a photovoltaic power generation performance evaluation device 200 in the first and second embodiments. DETAILED DESCRIPTION OF THE INVENTION

[0010] (First embodiment) Hereinafter, embodiments of the present disclosure will be described with reference to the drawings. The present invention is not limited to these embodiments. The drawings are schematic or conceptual, and the proportions of the various parts are not necessarily the same as those in reality. In the specification and drawings, elements similar to those described above with reference to the previous drawings are designated by the same reference numerals, and detailed descriptions thereof will be omitted as appropriate.

[0011] Additionally, in this disclosure, the terms "equal to or greater than" and "equal to or less than" can be read as "greater than" and "less than," respectively.

[0012] FIG. 1 is a schematic configuration diagram of a solar power plant 100 according to the first embodiment.

[0013] The solar power plant 100 in this embodiment includes a PV (Photovoltaic) panel 110, a junction box 120, a PCS (Power Conditioning System) package 131, extra-high voltage transformer equipment 144, a power measuring instrument 150, an environmental measuring instrument 160, and a solar power generation performance evaluation device 200.

[0014] For simplicity of explanation, this diagram shows an example in which multiple PV panels 110 are connected to one junction box 120, and the junction box 120 is further connected to one PCS package 131, but the configuration may be different depending on the scale of the solar power plant 100. For example, the solar power plant 100 may be configured to be provided with multiple junction boxes 120, and multiple PV panels 110 may be connected to each junction box 120 in units of a predetermined capacity. Alternatively, the solar power plant 100 may be configured to be provided with multiple PCS packages 131, and multiple junction boxes 120 may be connected to each PCS package 131 in units of a predetermined capacity.

[0015] The photovoltaic power generation performance evaluation device 200 evaluates the power generation performance of the photovoltaic power plant 100 using a learning model in which meteorological data is used as an explanatory variable and the results of simulating the power generation performance of the photovoltaic power plant 100 are used as a target variable. For example, the simulation uses on-premise dedicated software for determining a guaranteed power generation amount. In the simulation, based on the input of various meteorological data, the ideal power generation amount or PR (Performance Ratio) value under those conditions is output. Since the output power generation amount or PR value is the ideal output value under those meteorological conditions, these values ​​are also referred to as ideal power generation performance.

[0016] Furthermore, in this example, the solar power generation performance evaluation device 200 is installed in a building on the premises of the solar power plant 100, but the installation location of the device is not limited to this example, and it may be configured to be installed outside the premises of the solar power plant 100, such as in a data center, and connected via a network.

[0017] In this example, the PCS package 131 is disposed in the intermediate substation 130, and the extra-high voltage transforming equipment 144 is disposed in the extra-high voltage substation 140. The detailed configurations of these will be described later.

[0018] The power generated by the PV panels 110 is collected by a junction box 120 , and then boosted to a predetermined voltage at an intermediate substation 130 and an extra-high voltage substation 140 , and transmitted to a commercial grid 300 .

[0019] FIG. 2 is a block diagram showing the configuration of the solar power plant 100 in the first embodiment.

[0020] Fig. 2 is a block diagram showing the overall configuration of the photovoltaic power plant 100 of Fig. 1. Each block diagram of the photovoltaic power generation performance evaluation device 200 will be described later.

[0021] In this example, in the solar power plant 100, three PV panels 110 out of a plurality of PV panels 110 are connected to one junction box 120. Furthermore, a plurality of junction boxes 120 are connected to one PCS package 131, and a plurality of PCS packages 131 are connected to an extra-high voltage substation 140 in this unit.

[0022] The junction box 120 connects multiple PV panels 110 and aggregates the output power. The aggregated power is input to the PCS package 131. The number of connected PV panels 110 is not limited to three, and can be any number depending on the scale of the solar power plant 100, the capacity of the PCS 132, etc.

[0023] A plurality of PCS packages 131 are installed in the intermediate substation 130. Each PCS package 131 includes a PCS 132 and an intermediate transformer 133.

[0024] The PCS 132 is, for example, an inverter that converts a DC voltage input from the junction box 120 into an AC voltage. The intermediate transformer 133 converts the voltage output from the PCS 132 into a predetermined voltage.

[0025] The extra-high voltage substation 140 is equipped with a high-voltage switchgear 141, an extra-high voltage transformer 142, and an extra-high voltage switchgear 143.

[0026] The high-voltage switch gear 141 is a switch that switches whether or not the high-voltage current input from each intermediate transformer 133 is supplied to the extra-high voltage transformer 142. When the high-voltage switch gear 141 is in the on state, the high-voltage current is supplied to the extra-high voltage transformer 142.

[0027] The extra-high voltage transformer 142 converts the intermediate voltage input from each intermediate transformer 133 via the high voltage switch gear 141 into an extra-high voltage.

[0028] The extra-high voltage switch gear 143 switches whether or not to output the output power of the extra-high voltage transformer 142 to the commercial grid 300. When the extra-high voltage switch gear 143 is in the on state, the output power of the extra-high voltage transformer 142 is supplied to the commercial grid 300 from an interconnection point P, which is the output end of the extra-high voltage switch gear 143, as power generated by the photovoltaic power plant 100.

[0029] The power measuring device 150 measures the amount of power consumed by power devices such as the junction box 120 , the PCS 132 , the intermediate transformer 133 , the high-voltage switchgear 141 , the extra-high-voltage transformer 142 and the extra-high-voltage switchgear 143 .

[0030] The environmental measuring devices 160 are, for example, a pyranometer 161, a thermometer 162, and a wind condition meter 163. The pyranometer 161 measures the solar radiation intensity of sunlight incident on the PV panel 110. The thermometer 162 measures the temperature in the photovoltaic power plant. The wind condition meter 163 measures the wind force and wind direction in the photovoltaic power plant 100.

[0031] In this embodiment, an example will be described in which the photovoltaic power generation performance evaluation device 200 evaluates the performance of the photovoltaic power plant 100 based on meteorological data acquired from the above-mentioned environmental measuring device 160. However, the environmental measuring device 160 used for the performance evaluation is not limited to the above-mentioned example, and may be various measuring devices that measure the environment of the photovoltaic power plant 100. For example, the environmental measuring device 160 may be a hygrometer that measures the humidity of the photovoltaic power plant 100, a rain gauge that measures rainfall, a snow gauge that measures snowfall, a dirt meter that detects dirt, a wind vane and anemometer that measures wind direction and wind speed, or a thermometer that measures the temperature of the PV panel 110. In this embodiment, the data measured by the environmental measuring device 160 is treated as meteorological data.

[0032] The measurement data of the power measuring device 150 and the environmental measuring device 160 may be transmitted to the photovoltaic power generation performance evaluation device 200 via a network. The network is constructed using, for example, wired communication lines such as optical communication lines or metal communication lines, as well as wireless communication lines such as micro-wireless communication lines.

[0033] Hereinafter, the reference point in time, for example, the point in time at which power generation performance is evaluated, will be referred to as the second point in time, and a point in time prior to the second point in time will also be referred to as the first point in time. For ease of explanation, the weather data at the first point in time will also be referred to as past weather data.

[0034] FIG. 3 is a block diagram of a photovoltaic power generation performance evaluation device 200 in the first embodiment.

[0035] The photovoltaic power generation performance evaluation device 200 includes an explanatory variable preprocessing unit 250 , a learning model generation unit 260 , an acquisition unit 220 , a data processing unit 230 , an output unit 240 , and a storage unit 210 .

[0036] For example, past weather data of the solar power plant 100 can be used as the weather data to be the explanatory variables. However, there may be cases where no past weather data exists, such as immediately after the solar power plant 100 starts operating. The past weather data may be past weather data provided by government agencies, independent administrative agencies, private companies, etc., or dummy weather data. Of course, if past weather data obtained from the environmental measuring device 160 exists, this data may be used.

[0037] It is difficult to prepare all combinations of explanatory variables, such as meteorological data used to calculate the solar altitude, for example, the date (or information related to the solar position, such as the solar angle, the angle of incidence of sunlight on the panel, and the azimuth angle), solar radiation intensity, outside temperature, humidity, wind direction, wind speed, etc. Furthermore, simulations using many explanatory variables as inputs can be difficult due to the high human workload involved. The photovoltaic power generation performance evaluation device 200 of this embodiment analyzes statistical information, such as the maximum value, minimum value, mode, median, mean value, variance, and deviation of each meteorological data item that serves as an explanatory variable, comprehensively sets the range of values ​​that can be taken, and extracts or interpolates data to be used as an explanatory variable.

[0038] The photovoltaic power generation performance evaluation device 200 is, for example, a PC (Personal Computer) and has hardware resources such as a CPU (Central Processing Unit), ROM (Read Only Memory), RAM (Random Access Memory), HDD (Hard Disc Drive), and GPU (Graphics Processing Unit). The photovoltaic power generation performance evaluation device 200 can be realized by installing a program for the photovoltaic power generation performance evaluation device 200. The CPU in the photovoltaic power generation performance evaluation device 200 executes the program for the photovoltaic power generation performance evaluation device 200, thereby realizing the functions of a storage unit 210, an acquisition unit 220, a data processing unit 230, an output unit 240, an explanatory variable preprocessing unit 250, and a learning model generation unit 260. The storage unit 210 has, for example, an area for storing learning models and is constructed, for example, on an auxiliary storage device on an HDD.

[0039] The acquisition unit 220 acquires power generation amount data and weather data from the power measuring device 150 and the environmental measuring device 160. More specifically, the acquisition unit 220 acquires power generation amount data from the power measuring device 150 at a predetermined timing, and acquires weather data from the environmental measuring device 160. These data acquired by the acquisition unit 220 are stored in the storage unit 210. When acquiring past weather data from an external server of a government agency, an independent administrative agency, a private company, or the like, the acquisition unit 220 may acquire the data using, for example, a Web API (Application Programming Interface) or the like.

[0040] The explanatory variable preprocessing unit 250 performs preprocessing of the weather data that will be used as explanatory variables when generating a learning model. The explanatory variable preprocessing unit 250 analyzes statistical information such as the maximum value, minimum value, mode, median, mean value, variance, and deviation of the weather data that will be used as explanatory variables, comprehensively sets the range of values ​​that can be taken, and performs interpolation of the data to be used as explanatory variables to create explanatory variables.

[0041] The learning model generation unit 260 generates a learning model using weather data that serves as an explanatory variable. The learning model generation unit 260 inputs the weather data that serves as an explanatory variable into a simulation model to perform a simulation and calculates the power generation performance of the photovoltaic power plant 100 as a target variable. The learning model generation unit 260 generates a learning model by learning the response using AI (Artificial Intelligence), machine learning, or the like. By generating such a learning model, the photovoltaic power generation performance evaluation device 200 can calculate in real time the power generation performance of the photovoltaic power plant 100 under the weather conditions of the acquired weather data. In the simulation, the learning model generation unit 260 calculates the amount of power generation using a simulation model in which each device of the photovoltaic power plant 100 shown in FIG. 3 is modeled.

[0042] The learning model generation unit 260 also generates a learning model by machine learning the weather data that serves as input data and the power generation amount that serves as the simulation result. For machine learning, for example, a neural network (NN), a support vector machine (SMV), a decision tree, a random forest, a Bayesian method, or a logistic regression is used. These learning methods are merely examples, and other methods may also be used.

[0043] Furthermore, the learning model generation unit 260 stores the generated learning model in the storage unit 210.

[0044] There may be cases where no past weather data exists, such as immediately after the solar power plant starts operation. Therefore, as past weather data, weather data for the latitude and longitude where the solar power plant 100 is located or the surrounding area provided by a government agency, an independent administrative agency, a private company, or the like may be used. Furthermore, when latitude and longitude information is used as past weather data, altitude information may also be used. The surrounding area where the solar power plant 100 is located may be, for example, the administrative district where the solar power plant 100 is located.

[0045] Furthermore, instead of past weather data, dummy weather data may be used as explanatory variables. The dummy weather data is artificial weather data such as wind speed, wind direction, and temperature set within a range that can be taken depending on the simulation conditions.

[0046] In the following, the weather data that serves as learning data for the learning model will also be referred to as first weather data, and the weather data at the time of evaluating the power generation performance will also be referred to as second weather data. For example, past weather data and dummy weather data are examples of first weather data, and the weather data at the time of evaluating the power generation performance acquired by the acquisition unit 220 is an example of second weather data.

[0047] The data processing unit 230 evaluates the power generation performance of the photovoltaic power plant 100 using the power generation amount data and weather data acquired by the acquisition unit 220 at the time of evaluating the power generation performance. When evaluating the power generation performance, the data processing unit 230 inputs weather data into a trained learning model. The data processing unit 230 acquires the power generation amount or PR value output from the learning model based on the input weather data. Furthermore, the data processing unit 230 evaluates the power generation performance of the photovoltaic power plant 100 based on this result. The threshold value used in evaluating the power generation performance is also referred to as a first threshold value.

[0048] Furthermore, the data processing unit 230 performs a process of converting the acquired weather data at the time of evaluation of power generation performance into learning data so that the weather data can be applied to the learning model stored in the storage unit 210. Specifically, the data processing unit 230 performs, for example, input value range adjustment, input dimension adjustment, time resampling, outlier removal, and missing value processing on the weather data. The various conversion processes for applying weather data to a learning model are also called conversion processes.

[0049] The output unit 240 outputs the evaluation results acquired by the data processing unit 230. For example, the output unit 240 outputs, as performance evaluation results of the solar power plant 100, the power generation performance output by the learning model, the power generation performance of the solar power plant 100, the weather data used in the evaluation, and the accuracy of the learning model, for example, via a UI (User Interface) displayed on a display (not shown). Note that this display may be integrated with the PC or may be separate. Furthermore, the display may be provided in another PC connected via a network. In this case, the output unit 240 may control the output unit 240 provided in the other PC to output the evaluation results.

[0050] In addition, in this embodiment, an example in which the output unit 240 outputs the evaluation result to a display will be described, but other modes are also possible. For example, the output unit 240 may output the evaluation result to a projector other than a display, or may print the evaluation result on a paper medium using a printer.

[0051] The accuracy of the learning model may be determined, for example, by comparing the amount of power generated by the learning model with the amount of power generated at the time of evaluation. For example, if the error between the amount of power generated by the learning model and the amount of power generated at the time of evaluation is 10% or less, it can be determined that the difference is small, and the accuracy of the learning model can be determined as good. Otherwise, if the difference between these values ​​is large, the accuracy of the learning model can be determined as bad. Furthermore, if the output from the learning model is a PR value, the accuracy of the learning model may be determined, for example, by comparing the PR value output by the learning model with the PR value at the time of evaluation in the same manner as described above. The method of determining the accuracy of the learning model is one example, and methods other than the above method may also be used.

[0052] FIG. 4 is an example of weather data in the first embodiment.

[0053] This figure shows the plot of weather data from the past 30 years at a solar power plant 100, with the horizontal axis representing temperature (°C) and the vertical axis representing solar radiation intensity (W / m2). While this figure is presented as two-dimensional data, it may be expanded to N dimensions depending on the type of weather data. The graphs on the top and right sides of the figure show the frequency of occurrence of temperature and solar radiation intensity, respectively. As shown in the graphs on the top and right sides of the figure, there is a bias in the range of values, with the weather data concentrating on temperatures around 10-20°C and solar radiation intensity around 400-800 W / m2, indicating that the lower the temperature, the fewer samples there are. When performing a simulation using such weather data, the estimation error of the learning model may be large in areas with a low distribution of weather data.

[0054] In this embodiment, when there is variation in the distribution of weather data, the explanatory variable preprocessing unit 250 determines the range within which the weather data values ​​can be obtained, and then changes the number of pieces of weather data so as to reduce the distribution bias within this range. For example, the explanatory variable preprocessing unit 250 interpolates and thins out the weather data. After the explanatory variable preprocessing unit 250 performs a simulation using the interpolated and thinned weather data, the learning model generation unit 260 trains the response using AI, machine learning, or the like to generate a learning model.

[0055] FIG. 5 is an example of interpolation of meteorological data in the first embodiment.

[0056] The explanatory variable preprocessing unit 250 determines the range of possible weather data values ​​from the maximum and minimum values ​​of solar radiation intensity and temperature, and interpolates and thins out the weather data within that range. The interpolated data here is an example of dummy weather data. The plot in the figure shows how the explanatory variable preprocessing unit 250 interpolates the weather data within the possible data range to reduce bias in the distribution of the weather data. As a method for interpolating or thinning out the weather data, the explanatory variable preprocessing unit 250 may use various statistical methods, such as mean value substitution, regression substitution, ratio substitution, or Hotdeck. While it is preferable to interpolate the weather data so that the values ​​are evenly spaced within the possible range, the explanatory variable preprocessing unit 250 may also determine the interpolation interval based on the distribution of the weather data.

[0057] Furthermore, when highly accurate estimation is desired under certain conditions, the explanatory variable preprocessing unit 250 may interpolate weather data unevenly within the entire range according to the deviation of the distribution. By narrowing the range of weather data to be interpolated unevenly rather than uniformly across the entire range, the explanatory variable preprocessing unit 250 may reduce the accuracy of the learning model for input weather data under conditions that occur infrequently, but may improve the estimation accuracy under certain conditions.

[0058] FIG. 6 shows an example in which the range of possible values ​​of the weather data in the first embodiment is expanded.

[0059] The explanatory variable preprocessing unit 250 may expand the range of meteorological data values ​​by assigning likelihood to the maximum and minimum values. In this figure, the explanatory variable preprocessing unit 250 assigns a likelihood of -β to the minimum temperature value and a likelihood of +β to the maximum temperature value. In addition, the explanatory variable preprocessing unit 250 assigns a likelihood of +α to the maximum solar radiation intensity value.

[0060] For example, this is effective when the amount of meteorological data that is the source of the learning data is small or when the user does not understand the conditions that may occur.

[0061] FIG. 7 shows an example of excluding weather data in the first embodiment.

[0062] The explanatory variable preprocessing unit 250 excludes meteorological data that is not used as an explanatory variable. The gray area in the figure shows an example of the range of meteorological data that the explanatory variable preprocessing unit 250 excludes, that is, the range of meteorological data that is not used as an explanatory variable. When a user wants to check the power generation performance of the solar power plant 100 under specific conditions, by limiting the meteorological data used as an explanatory variable to a specific range, the solar power generation performance evaluation device 200 can create a learning model that enables estimation with higher accuracy.

[0063] For example, the range of meteorological data to be used as explanatory variables may be input by a user using an input device (not shown), such as a mouse or keyboard, via a UI displayed on a display (not shown). The user may input desired values ​​for solar radiation intensity and temperature, and the explanatory variable preprocessing unit 250 may exclude meteorological data based on these values. Alternatively, the explanatory variable preprocessing unit 250 may automatically exclude meteorological data without requiring user input of values. The example in FIG. 7 shows an example in which data with temperatures below -10°C and data with solar radiation intensity below 100 W / m2 are excluded.

[0064] The learning model generation unit 260 performs a simulation using, for example, the weather data that has been subjected to the preprocessing shown in FIGS. 5 to 7 described above, and generates a learning model by learning the response through machine learning.

[0065] FIG. 8 is an example of a flowchart for generating a learning model in the first embodiment.

[0066] In this flowchart, the photovoltaic power generation performance evaluation device 200 generates a learning model using past weather data. For simplicity of explanation, this flowchart describes an example in which the photovoltaic power generation performance evaluation device 200 generates a learning model using past weather data at a single point in time, but the learning model may also be generated using past weather data from several years to several decades, for example.

[0067] In step S1, the explanatory variable preprocessing unit 250 acquires past weather data from the storage unit 210. The acquired past weather data may be data from a single time point or multiple time points. In step S2, the explanatory variable preprocessing unit 250 performs preprocessing of the explanatory variables on the acquired past weather data. As shown in FIG. 5, the explanatory variable preprocessing unit 250 may preprocess the explanatory variables by determining a possible range of weather data and interpolating and thinning out weather data within that range. As shown in FIG. 6, the explanatory variable preprocessing unit 250 may preprocess the explanatory variables by expanding the possible range of weather data by assigning likelihood to maximum and minimum values. As shown in FIG. 7, the explanatory variable preprocessing unit 250 may preprocess the explanatory variables by excluding weather data outside the range used as weather data.

[0068] In step S3, the learning model generation unit 260 performs a simulation using, for example, preprocessed weather data as input data for the simulation model. In the simulation, a simulation model that models each device of the solar power plant 100 shown in FIG. 3 is used. In step S4, the learning model generation unit 260 calculates ideal power generation performance through simulation. For example, the learning model generation unit 260 calculates the amount of power generation or PR value under the input weather conditions. In addition to the weather data input in step S3, the amount of power generation or PR value calculated in this step is used as learning data for the learning model. When past weather data from multiple points in time is used, simulations are performed under each weather condition.

[0069] In step S5, the learning model generation unit 260 learns the response of the simulation model to the preprocessed weather data by machine learning, and generates a learning model. The learning model generation unit 260 also stores the generated learning model in the storage unit 210.

[0070] When generating a learning model, if the solar altitude angle is used as one of the explanatory variables, the solar altitude angle needs to be corrected every four years to eliminate discrepancies, taking leap years into account. Therefore, the learning model generation unit 260 may generate a learning model by learning the response of the simulation model to preprocessed weather data every four years through machine learning.

[0071] FIG. 9 is an example of a flowchart for evaluating power generation performance in the first embodiment.

[0072] This flowchart describes an example in which the photovoltaic power generation performance evaluation device 200 acquires weather data and power generation amount data, and then evaluates the power generation performance of the photovoltaic power plant 100 to be evaluated. In addition, this flowchart will be described assuming that the learning model generated by the learning model generation unit 260 is stored in the storage unit 210.

[0073] In step S21, the acquisition unit 220 acquires weather data at the time of evaluation from the environmental measuring device 160. The acquisition unit 220 also stores this weather data in the storage unit 210. The stored weather data is used to evaluate power generation performance as well as to generate and update a learning model. In step S22, the data processing unit 230 performs conversion processing so that the acquired weather data can be applied to the learning model. The data processing unit 230 performs, for example, input value range matching, input dimension matching, time resampling, outlier removal, and missing value processing.

[0074] In step S23, the data processing unit 230 inputs the converted weather data into the learning model. This weather data is used as explanatory variables of the learning model. In step S24, the learning model outputs the power generation amount or PR value corresponding to the input weather data. The power generation amount or PR value output by the learning model using the weather data measured at the photovoltaic power plant 100 at the time of evaluation of the power generation performance as input is also referred to as system power generation performance. The system power generation performance is determined between the power generation business operator and the electric power company, and may be used as a guaranteed power generation amount, which is the expected power generation amount under certain weather conditions.

[0075] In step S25, the data processing unit 230 derives the power generation amount at the interconnection point P using the acquired power generation amount data. The power generation amount at the interconnection point P indicates the total power generation amount of the photovoltaic power plant 100 measured by the power metering device 150. The data processing unit 230 may, for example, extract the amount of power measured at the interconnection point P from the power generation amount data of the power metering device 150. The amount of power or the PR value measured at the interconnection point P at the time of evaluating the power generation performance is also referred to as actual power generation performance. In step S26, the data processing unit 230 compares the actual power generation performance with the system power generation performance to determine the power generation performance. If the difference between the actual power generation performance and the system power generation performance as a result of comparing the two is equal to or greater than a predetermined first threshold, the data processing unit 230 determines that some kind of malfunction, such as an abnormality or deterioration, has occurred in the photovoltaic power plant 100. The comparison between the actual power generation performance and the system power generation performance may be performed using a method other than using the difference between these values.

[0076] In step S27, the output unit 240 outputs the determination result by the data processing unit. The output unit 240 may output information indicating the performance of the solar power plant 100, for example, by using text information such as "No abnormality" or "Abnormality present" regarding the presence or absence of an abnormality in the solar power plant 100. The output unit 240 may also output the system power generation performance, the actual power generation performance, the weather data used, and the accuracy of the learning model. The accuracy of the learning model may be calculated by comparing the system power generation performance with the actual power generation performance. For example, when the system power generation performance is 98 MWh and the actual power generation performance is 100 MWh, if the error between the actual power generation amount and the system power generation performance is less than a first threshold (e.g., less than 10%), the accuracy of the learning model may be considered good. If the error is equal to or greater than the first threshold, the accuracy of the learning model may be considered bad.

[0077] Furthermore, the output unit 240 may notify the status of the solar power plant 100 via email or a social networking service (SNS) to a communication terminal used by the manager of the solar power plant 100 or a user of the solar power generation performance evaluation device 200. If a malfunction occurs in the solar power plant 100, the manager and the user are notified at an early stage, so that the manager can grasp the actual power generation performance of the solar power plant 100 in real time.

[0078] According to this embodiment, the photovoltaic power generation performance evaluation device 200 performs preprocessing of the weather data that serves as the explanatory variables using the explanatory variable preprocessing unit 250. This reduces the burden on the user of creating a learning model, and enables more accurate performance evaluation of the photovoltaic power plant 100.

[0079] Furthermore, according to this embodiment, the photovoltaic power generation performance evaluation device 200 evaluates the power generation performance of the photovoltaic power plant 100 by comparing the system power generation performance output from the learning model using weather data at the time of evaluation of the power generation performance as input with the actual power generation performance derived from the power generation amount data of the power measuring device 150. It is possible to evaluate the power generation performance of the photovoltaic power plant 100 without using past weather data.

[0080] Furthermore, according to this embodiment, even when no past weather data exists, such as immediately after the solar power plant 100 starts operation, the power generation performance can be evaluated using dummy weather data that has been preprocessed by the explanatory variable preprocessing unit 250.

[0081] Furthermore, according to this embodiment, the photovoltaic power generation performance evaluation device 200 performs preprocessing to impart likelihood to a range of the maximum and minimum values ​​of the weather data using the explanatory variable preprocessing unit 250. This makes it possible to expand the range of weather data, and therefore, even when the amount of weather data is small or when the user does not understand the conditions that may occur, it is possible to change the range of weather data and evaluate power generation performance.

[0082] Furthermore, according to this embodiment, the photovoltaic power generation performance evaluation device 200 performs preprocessing to limit the range used as weather data to a specific range using the explanatory variable preprocessing unit 250. This enables the photovoltaic power generation performance evaluation device 200 to create a learning model that can estimate the power generation performance of the photovoltaic power plant 100 under specific conditions with higher accuracy.

[0083] In addition, since the solar power generation performance evaluation method uses a learning model to evaluate power generation performance, there is no need to calculate the amount of power generated using on-premise dedicated software, which shortens the time required to calculate the amount of power generated.

[0084] (Second embodiment) FIG. 10 is an example of a flowchart when updating a learning model in the second embodiment.

[0085] In this embodiment, explanations of parts similar to those in the first embodiment, such as a block diagram of the photovoltaic power generation performance evaluation device 200, will be omitted. The learning model before update is assumed to be stored in the storage unit 210. In this embodiment, the threshold value for determining whether or not to update the learning model is also referred to as the second threshold value.

[0086] When there are defects in the weather data or when there is a shortage of learning data, the accuracy of the learning model may not improve even if the learning model is updated using the weather data acquired by the acquisition unit 220. Therefore, the photovoltaic power generation performance evaluation device 200 in this embodiment compares the actual power generation performance with the system power generation performance, and if these values ​​are far apart, determines that the data does not improve the accuracy of the learning model, and does not use the weather data used at this time to update the learning model.

[0087] In the flowchart of this embodiment, the learning model generation unit 260 determines whether to update the learning model based on the results of a comparison between the actual power generation performance and the system power generation performance. For example, when the learning model generation unit 260 compares the two performances and the difference between these values ​​is equal to or greater than a second threshold, the weather data used to calculate the actual power generation performance at this time is determined to be an abnormal value and is not incorporated into the learning model. On the other hand, when the difference between these values ​​is smaller than the second threshold, the weather data at this time is determined to be a normal value and is incorporated into the learning model. In other words, the learning model generation unit 260 performs a simulation using this weather data and updates the learning model by learning the response using machine learning.

[0088] In this example, the decision as to whether to update the learning model is made based on whether the difference between the actual power generation performance and the system power generation performance is within a specified value, but this is not limited to this method and various other decision methods may be used.

[0089] The operations from step S31 to step S37 are the same as the operations from step S21 to step S27, and therefore the explanation will be omitted.

[0090] In step S38, the explanatory variable preprocessing unit 250 calculates the difference between the actual power generation performance and the system power generation performance, and determines whether this difference is equal to or greater than a predetermined second threshold. If the difference between the actual power generation performance and the system power generation performance is equal to or greater than the second threshold (YES in step S38), the learning model generation unit 260 does not update the learning model and ends the process. If the difference between the actual power generation performance and the system power generation performance is smaller than the second threshold (NO in step S38), the learning model generation unit 260 inputs the weather data used for the actual power generation performance into the simulation.

[0091] In step S40, the learning model generation unit 260 calculates ideal power generation performance through simulation. For example, the learning model generation unit 260 calculates the power generation amount or PR value under the input weather conditions. In step S41, the learning model generation unit 260 learns the response of the simulation model through machine learning and updates the learning model. In addition, the learning model generation unit 260 stores the updated learning model in the storage unit 210.

[0092] According to this embodiment, the photovoltaic power generation performance evaluation device 200 determines whether the meteorological data updates the learning model by comparing the actual power generation performance with the system power generation performance. As a result, only the data suitable for improving the accuracy of the learning model can be used for updating among the meteorological data acquired by the acquisition unit 220.

[0093] FIG. 11 is a hardware configuration diagram of the photovoltaic power generation performance evaluation device 200 in the first and second embodiments.

[0094] 11 includes a processor 52 such as a CPU, a main memory device 53 such as a RAM, an auxiliary memory device 54 such as an HDD, a network interface 55 such as a LAN (Local Area Network) board, a device interface 56 such as a memory slot or memory port, and a bus 57 that connects these devices to each other. The solar cell diagnosis device 10 is, for example, a computer such as a PC, and includes external input devices such as a keyboard and a mouse, and an output device such as an LCD monitor.

[0095] In this embodiment, a program for causing a computer to execute information processing of the photovoltaic power generation performance evaluation device 200 is installed in the auxiliary storage device 54. The photovoltaic power generation performance evaluation device 200 loads this program into the main storage device 53 and executes it using the processor 52. As a result, the functions of the storage unit 210, acquisition unit 220, data processing unit 230, output unit 240, explanatory variable preprocessing unit 250, and learning model generation unit 260 shown in FIG. 3 are realized in the photovoltaic power generation performance evaluation device 200, and performance evaluation of the photovoltaic power plant 100 described in the first embodiment becomes possible. Note that data generated by this information processing is temporarily held in the main storage device 53 or stored and saved in the auxiliary storage device 54.

[0096] Furthermore, the storage unit 210 is constructed on the auxiliary storage device 54. The above-mentioned first and second thresholds are stored in the auxiliary storage device 54. The thresholds are loaded into the main storage device 53 when this program is executed.

[0097] The photovoltaic power generation performance evaluation device 200 is also connected to the network 7 via a network interface 55. The photovoltaic power generation performance evaluation device 200 controls the network interface 55 by an acquisition unit 220 to acquire weather data and power generation amount data.

[0098] The power generation performance evaluation program for the photovoltaic power generation performance evaluation device 200 can be installed, for example, by attaching an external device 58 recording the program to the device interface 56 and storing the program from the external device 58 in the auxiliary storage device 54. Examples of the external device 58 include a computer-readable recording medium and a recording device incorporating such a recording medium. Examples of recording media include a CD-ROM (Compact Disk Read Only Memory), a CD-R (Compact Disk Recordable), a flexible disk, a DVD-ROM (Digital Versatile Disk Read Only Memory), and a DVD-R (Digital Versatile Disk Recordable), and an example of a recording device is a HDD. Also, the program can be installed, for example, by downloading the program via the network interface 55.

[0099] According to this embodiment, the functions of the photovoltaic power generation performance evaluation device 200 in any one of the first and second embodiments can be realized by software.

[0100] Although several embodiments have been described above, these embodiments are presented only as examples and are not intended to limit the scope of the invention. The novel photovoltaic power generation performance evaluation device 200 described in this specification can be embodied in various other forms. Furthermore, various omissions, substitutions, and modifications can be made to the form of the photovoltaic power generation performance evaluation device 200 described in this specification without departing from the spirit of the invention. The appended claims and their equivalents are intended to include such forms and modifications that fall within the scope and spirit of the invention. [Explanation of symbols]

[0101] 52: Processor, 53: Main memory device, 54: Auxiliary memory device, 55: Network interface, 56: Device interface, 57: Bus, 58: external device, 100: solar power plant, 110: PV panel, 120: connection box, 130: Intermediate substation, 131: PCS package, 132: PCS, 133: Intermediate transformer, 140: Extra high voltage substation, 141: High voltage switchgear, 142: Extra-high voltage transformer, 143: Extra-high voltage switchgear, 144: Extra-high voltage substation equipment, 150: Power measuring equipment, 160: Environmental measuring equipment, 161: Pyranometer, 162: Thermometer, 163: Wind condition meter, 200: Photovoltaic power generation performance evaluation device, 210: Memory unit, 220: Acquisition unit, 230: Data processing unit, 240: Output unit, 250: explanatory variable preprocessing unit, 260: learning model generation unit, 300 commercial power system

Claims

1. an explanatory variable preprocessing unit that preprocesses first weather data, which is weather data to be used as learning data; a learning model generation unit that generates a learning model by machine learning using the first weather data as an explanatory variable and a result of simulating the system power generation performance in the solar power plant as a target variable; an acquisition unit that acquires second weather data, which is the weather data at the time of evaluating the power generation performance, and power generation amount data in the photovoltaic power plant; a data processing unit that compares the system power generation performance output from the learning model in response to input of the second weather data with an actual power generation performance derived from the power generation amount data, and evaluates the power generation performance of the solar power plant. Photovoltaic power generation performance evaluation device.

2. The explanatory variable preprocessing unit determining a possible range for the first weather data; performing a process of changing the number of the first weather data so as to reduce the bias in distribution within a range that can be taken as the first weather data; The photovoltaic power generation performance evaluation device according to claim 1 .

3. The photovoltaic power generation performance evaluation device according to claim 2 , wherein the explanatory variable preprocessing unit performs a process of interpolating or thinning out the first weather data within a range that can be taken as the first weather data.

4. The photovoltaic power generation performance evaluation device according to claim 1 , wherein the explanatory variable preprocessing unit performs processing to give likelihood to the maximum and minimum values ​​of the first weather data, thereby expanding the range of values ​​that can be taken.

5. The photovoltaic power generation performance evaluation device according to claim 1 , wherein the explanatory variable preprocessing unit performs processing to exclude the first meteorological data in a range that is not used as an explanatory variable.

6. The photovoltaic power generation performance evaluation device according to claim 3 , wherein the explanatory variable preprocessing unit performs a process of non-uniformly interpolating the first weather data within a possible range of the first weather data.

7. 2 . The photovoltaic power generation performance evaluation device according to claim 1 , wherein the learning model generation unit learns a response of a simulation model to the first weather data every four years by machine learning, and generates the learning model.

8. the explanatory variable preprocessing unit compares the actual power generation performance with the system power generation performance; The learning model generation unit updates the learning model based on a result of the comparison. The photovoltaic power generation performance evaluation device according to claim 1 .

9. 2. The photovoltaic power generation performance evaluation device according to claim 1, wherein the data processing unit determines that a malfunction has occurred in the photovoltaic power plant when a difference between the actual power generation performance and the system power generation performance is equal to or greater than a predetermined first threshold value.

10. 9. The solar power generation performance evaluation device according to claim 8, wherein the explanatory variable preprocessing unit compares the actual power generation performance with the system power generation performance, and if the difference between the actual power generation performance and the system power generation performance is equal to or greater than a predetermined second threshold, determines not to update the learning model.

11. Preprocessing first weather data, which is weather data to be used as learning data; generating a learning model by machine learning using the first weather data as an explanatory variable and a result of simulating the system power generation performance of the solar power plant as a target variable; acquiring second weather data, which is the weather data at the time of evaluating the power generation performance, and power generation amount data at the photovoltaic power plant; comparing the system power generation performance output from the learning model in response to the input of the second weather data with an actual power generation performance derived from the power generation amount data, and evaluating the power generation performance of the solar power plant; A photovoltaic power generation performance evaluation method including:

12. Preprocessing first weather data, which is weather data to be used as learning data; generating a learning model by machine learning using the first weather data as an explanatory variable and a result of simulating the system power generation performance of the solar power plant as a target variable; acquiring second weather data, which is the weather data at the time of evaluating the power generation performance, and power generation amount data at the photovoltaic power plant; comparing the system power generation performance output from the learning model in response to the input of the second weather data with an actual power generation performance derived from the power generation amount data, and evaluating the power generation performance of the solar power plant; A photovoltaic power generation performance evaluation program that causes a computer to execute a photovoltaic power generation performance evaluation method including the steps of:

Images