Information processing device, information processing system, and information processing method

Abstract

Improves the efficiency of operation and management of multiple power sources. [Solution] Actual measurement data showing the actual measured value of the amount of electricity supplied is obtained for each power source, and for each power source, a power source where the deviation between the predicted value of the amount of electricity supplied, calculated using a predetermined mathematical model based on weather data showing the weather at the installation location of the power source, and the actual measured value is equal to or greater than a predetermined reference value, is selected as a candidate power source to be investigated, inquiry information is generated to inquire about the cause of any abnormality in the amount of electricity supplied, including the actual measured value and the predicted value of the candidate power source, the inquiry information is output to a predetermined analytical model, and response information including the cause of the abnormality and the need for investigation is obtained from the analytical model.

Description

[Technical Field]

[0001] The present application relates to an information processing device, an information processing system, and an information processing method, for example, an information processing system for supporting the maintenance management of a power supply. [Background technology]

[0002] With growing interest in environmental issues, the use of renewable energy sources with low environmental impact is being promoted. In response to this trend, power sources using renewable energy are becoming more widespread. Examples of such power sources include power generation facilities such as solar power plants and wind power plants. Furthermore, operational management of power generation facilities is important for ensuring a stable power supply. Therefore, methods for determining the status of power generation facilities have been proposed.

[0003] For example, the diagnostic device for a solar power generation facility described in Patent Document 1 includes a first measurement unit that measures the value of output power or the value of output current in the solar power generation facility, an estimation unit that estimates the cause of a decrease in power generation amount based on time-dependent data linking the measured value of output power or the value of output current with the measurement time and a predetermined coefficient, and a learning unit that updates the coefficient using teacher data, which is time-dependent data in which the cause of the decrease in power generation amount is specified in advance.

[0004] The diagnostic device described in Patent Document 2 includes a first feature data generation unit that generates multiple first feature data, which are feature data representing the characteristics of multiple electrical physical quantities generated by multiple strings or multiple power generation modules during a specified learning period; a division unit that time-divides the multiple first feature data to generate learning data; a second feature data generation unit that generates second feature data, which is feature data at the measurement time of the diagnostic target; a learning unit that learns the learning data using a one-class support vector machine; a judgment unit that judges whether the second feature data is an outlier based on the boundary determined by the learning; and a diagnostic unit that diagnoses the power generation system based on the judgment result.

[0005] The power amount evaluation device described in Patent Document 3 includes a calculation unit that calculates differential power, which is the difference between the actual power amount during an operation period, the power amount estimated from actual data during the operation period, and the power amount predicted for the operation period, for each of a plurality of periods, and estimates the difference factor of the differential power for each of the plurality of periods, and a memory unit that displays the differential power and the difference factor calculated by the calculation unit on a display unit. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Publication No. 2019-208350 [Patent Document 2] Japanese Patent Publication No. 2022-057961 [Patent Document 3] Japanese Patent Application Publication No. 2024-127002 Summary of the Invention [Problem to be solved by the invention]

[0007] However, renewable energy power sources tend to be smaller in scale and more numerous and geographically dispersed than power sources that have a greater environmental impact, such as thermal power plants and nuclear power plants. Therefore, there is a need to improve the efficiency of the operation and management of these power sources. [Means for solving the problem]

[0008] The present application has been made to solve the above-mentioned problems, and an information processing device according to one aspect acquires actual measurement data indicating an actual measurement value of the amount of power supplied for each power source, and calculates a predicted value of the amount of power supplied, calculated for each power source using a predetermined mathematical model based on weather data indicating the weather at the installation location of the power source, and calculates a deviation between the predicted value and the actual measurement value, which is equal to or greater than a predetermined reference value. The aforementionedA power source is selected as a candidate power source to be investigated, inquiry information is generated to inquire about the cause of the abnormality in the supply amount, including the actual measured value and the predicted value of the candidate power source, the inquiry information is output to a predetermined analytical model, and response information including the cause of the abnormality and the need for investigation is obtained from the analytical model.

[0009] An information processing system according to one embodiment may include the above-described information processing device and a model processing unit that uses the above-described analytical model to estimate the cause of the abnormality and the need for an investigation of the abnormality from the inquiry information, and outputs response information including the cause and the need for an investigation to the information processing device.

[0010] An information processing method according to one aspect includes acquiring actual measurement data indicating an actual measurement value of an amount of power supply for each power source, and calculating a deviation between a predicted value of an amount of power supply calculated for each power source using a predetermined mathematical model based on weather data indicating weather at an installation location of the power source and the actual measurement value, the deviation being equal to or greater than a predetermined reference value. The aforementioned A power source is selected as a candidate power source to be investigated, inquiry information is generated to inquire about the cause of the abnormality in the supply amount, including the actual measured value and the predicted value of the candidate power source, the inquiry information is output to a predetermined analytical model, and response information including the cause of the abnormality and the need for investigation is obtained from the analytical model. [Effects of the Invention]

[0011] According to the embodiment of the present application, it is possible to improve the efficiency of power supply operation management. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a schematic block diagram illustrating an example of the configuration of an information processing system according to an embodiment of the present invention. [Figure 2] FIG. 2 is a schematic block diagram illustrating an example of the functional configuration of a terminal device according to the present embodiment. [Figure 3] FIG. 2 is a schematic block diagram illustrating an example of the functional configuration of an application server according to the present embodiment. [Figure 4]10 is a flowchart illustrating an analysis process according to the present embodiment. [Figure 5] 10A and 10B are diagrams illustrating actual measurement data and predicted data according to the present embodiment. [Figure 6] FIG. 2 is a diagram illustrating power plant data according to the present embodiment. [Figure 7] FIG. 2 is a diagram illustrating a power plant list according to the present embodiment. [Figure 8] FIG. 3 is a diagram showing a first example of survey information according to the present embodiment. [Figure 9] FIG. 10 is a diagram showing a second example of survey information according to the present embodiment. [Figure 10] FIG. 2 is a diagram illustrating an example of a schematic configuration of a prompt according to the present embodiment. [Figure 11] FIG. 10 is a diagram showing a specific example of a prompt according to the embodiment. [Figure 12] FIG. 4 is a diagram showing a first example of answer information according to the present embodiment. [Figure 13] FIG. 10 is a diagram showing a second example of answer information according to the present embodiment. [Figure 14] 10 is a sequence chart illustrating a registration process of survey information according to the present embodiment. [Figure 15] FIG. 2 is a diagram showing a first example of a display screen according to the present embodiment. [Figure 16] FIG. 10 is a diagram showing a second example of a display screen according to the present embodiment. [Figure 17] 10 is a sequence chart illustrating an analysis process according to the present embodiment. [Figure 18] FIG. 1 is a schematic block diagram illustrating an example of the configuration of a computer system according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0013] Hereinafter, an embodiment of the present application will be described with reference to the drawings. An example of the configuration of an information processing system S1 according to this embodiment will be described. FIG. 1 is a schematic block diagram showing an example of the configuration of an information processing system S1 according to this embodiment. The information processing system S1 includes a terminal device 10, an application server 20, a database server 30, and an AI server 40. The terminal device 10, the application server 20, the database server 30, and the AI ​​server 40 are communicably connected to one another via a network NW.

[0014] The network NW is also connected to communication devices provided in each of N power plants 50 (N is a predetermined integer equal to or greater than 1, typically 100 to 1000 or more). The N power plants 50 are, for example, power sources owned or managed by a specific power generation company. The individual power plants 50 are distinguished by being assigned sub-numbers such as power plant 50-1, -1, etc. The individual power plants are, for example, solar power plants. Each solar power plant has a solar power generation system (not shown) and is an example of a power source that supplies generated power to the power grid (not shown) by connecting to the power grid. The power plant 50 is equipped with a watt-hour meter (not shown) and a communication device (not shown). The watt-hour meter measures the amount of power generated by the generator, i.e., the amount of power generated. The communication device transmits actual measurement data indicating the amount of power measured by the watt-hour meter as an actual value of the amount of power supplied to the database server 30 via the network NW at predetermined intervals (for example, every 15 to 60 minutes, typically every 30 minutes). In this application, this period may be referred to as a "coma."

[0015] The power plant 50 may have a second watt-hour meter at a node with the power grid. The second watt-hour meter measures the amount of power transmitted from the power source to the power grid, i.e., the amount of power transmitted. The communication device may transmit actual measurement data indicating the amount of power measured by the second watt-hour meter as an actual value of the amount of power supplied to the database server at predetermined intervals, instead of or in addition to the amount of power generated. The power grid may include a power control system (not shown). The power control system is connected to individual power sources (such as power plant 50-1 in the example of FIG. 1) using a network NW. The power sources may be connected to the power control system (not shown) using communication devices, and the amount of power transmitted may be controlled according to commands from the power control system. The network NW includes, for example, a wide area network such as the Internet, a public communication network, etc. The network NW may be combined with a narrow area network such as a virtual private network, a local area network, etc.

[0016] The terminal device 10 provides a user interface for the application server 20. That is, the terminal device 10 accepts user operations and transmits various types of input information to the application server 20 based on the accepted operations. The terminal device 10 receives various types of received information from the application server 20 and presents a display screen showing output information based on the received information. Specific examples of the input information and output information will be described later. The terminal device 10 may be realized in any form, such as a personal computer (PC), a mobile phone, or a tablet terminal device.

[0017] The application server 20 is a server device 20 that cooperates with the database server 30 and the AI ​​server 40 to execute the analysis processing according to this embodiment. The application server 20 may be configured as, for example, a web server. When receiving input information from the terminal device 10, the application server 20 acquires actual measurement data and weather data for each power source from the database server 30. The actual measurement data is data that indicates the actual measurement value of the amount of power supplied by each power source. The weather data is data that indicates the weather at the installation location of each power source. The application server 20 calculates a predicted value of the supply amount using a predetermined mathematical model based on the weather data acquired for each power source. The application server 20 selects as candidate power sources those power sources for which the deviation between the actual measured value and the predicted value is equal to or greater than the predetermined reference value of the deviation degree. The candidate power sources correspond to power sources that are candidates for the on-site survey.

[0018] The application server 20 generates inquiry information for inquiring about the cause of the abnormality that occurred in the amount of power supplied by the candidate power source. The application server 20 transmits the generated inquiry information to the AI ​​server 40. The application server 20 receives response information from the AI ​​server 40 in response to the inquiry information. The response information is information indicating the cause of the abnormality that occurred in the actual measurement value of the supply amount and the need for an investigation. The application server 20 transmits to the terminal device 10 output information indicating the analysis result of the supply amount indicated in the response information.

[0019] The database server 30 is a server device that receives actual measured values ​​of the amount of power generated for each power plant 50 and stores measured data indicating the received measured values ​​of the amount of power generated. The stored measured data may include the amount of power transmitted. The database server 30 also receives weather data indicating the weather conditions at the location of each power plant 50 from a weather information providing system (not shown) and stores the received weather data. The weather information providing system is managed by, for example, a weather service provider, a government agency, a local government, or a local public organization that provides weather information. The database server 30 reads one or both of the actual measurement data and the meteorological data in response to a request from the application server 20 , and transmits the read data to the application server 20 .

[0020] The AI ​​(Artificial Intelligence) server 40 is a server device that generates a response to a query using a preset AI model. The AI ​​server 40 is configured, for example, as a web server. That is, the AI ​​server 40 executes a process of generating response information including the cause of an abnormality in the amount of power supplied by a candidate power source indicated in the query and the need to investigate that power source. The AI ​​server 40 uses a large language model (LLM) as the AI ​​model. The LLM is, for example, a transformer. The transformer includes an encoder and a decoder. The encoder calculates an intermediate token sequence indicating context information from an input token sequence indicating the query information. The decoder calculates an output token sequence indicating the response information from the intermediate token sequence obtained from the encoder. Through this calculation process, the response information is generated. The response information is not necessarily limited to text and may include images as part of it. The AI ​​server 40 transmits the generated response information to the application server 20 as a response to the query information.

[0021] Next, an example of the functional configuration of the terminal device 10 according to this embodiment will be described. Fig. 2 is a schematic block diagram showing an example of the functional configuration of the terminal device 10 according to this embodiment. The terminal device 10 includes a user interface unit 102 , an input device 112 , and a display 114 .

[0022] The user interface unit 102 realizes a function of providing a user interface to the application server 20. The host system of the terminal device 10 realizes the function of the user interface unit 102, for example, by executing a predetermined application program (sometimes referred to herein as an "app" or "application"). The host system of the terminal device 10 may realize the function of the user interface unit 102 by executing an application provided by the application server 20 on a browser. Note that, in this application, executing processes instructed by various commands written in a program (including an application, a browser, or other types of program) may be referred to as "executing a program" or "executing a program."

[0023] The user interface unit 102 generates input information for the application server 20 in response to instructions (i.e., user operations) notified by operation signals input from the input device 112. The user interface unit 102 transmits the generated input information to the application server 20. For example, the user interface unit 102 identifies a period indicated by the operation signal as a target period for analyzing the operating status of the power plant 50. The user interface unit 102 transmits input information indicating the identified target period to the application server 20.

[0024] The user interface unit 102 generates a display screen having a predetermined configuration, and outputs display data showing the generated display screen to the display 114. The display screen may be provided with input fields or screen components for inputting various input information, such as a target period, in response to user operations, and a display field for displaying output information to be provided. When receiving output information from the application server 20, the user interface unit 102 acquires display element information to be displayed based on the output information and arranges the acquired display element information on the display screen. The user interface unit 102 displays the display screen on which the display element information is arranged on the display 114. For example, the user interface unit 102 converts received information indicating an event that occurred when a significant difference appears between the actual measured value and the predicted value of the amount of power supplied from the power plant 50 as a result of analyzing the operating status of the power plant 50, the presumed cause, and the degree of necessity for an on-site investigation into output information in a format more suitable for viewing. The user interface unit 102 arranges the converted output information to construct a display screen and outputs display data indicating the constructed display screen to the display 114.

[0025] The input device 112 receives a user operation, generates an operation signal that transmits input information instructed in accordance with the received user operation, and outputs the generated operation signal to the user interface unit 102. The input device 112 may be, for example, a general-purpose component such as a touch sensor or a mouse, or may be a dedicated component such as a button or a dial. The display 114 displays a display screen indicated by display data input from the user interface unit 102. The display 114 may have any type of display medium, such as a liquid crystal display or an organic light-emitting diode display. The touch sensor constituting the input device 112 and the display panel constituting the display 114 may be integrated into one unit to form a touch panel.

[0026] Next, an example of the functional configuration of the application server 20 according to this embodiment will be described. FIG. 3 is a schematic block diagram showing an example of the functional configuration of the application server 20 according to this embodiment. The application server 20 includes an application processing unit 202 , an output processing unit 204 , an evaluation unit 206 , and a storage unit 208 .

[0027] In response to receiving input information from the terminal device 10, the application processing unit 202 generates inquiry information for inquiring about the cause of an abnormality in the supply amount for the power plant 50 that is a candidate power source and is a candidate to be investigated. The inquiry information is configured as a prompt, including the actual measured value of the amount of power supplied from the power plant 50 that is a candidate power source and is a candidate to be investigated, and the predicted value of the amount of power supplied based on the weather at the power plant 50.

[0028] The application processing unit 202 transmits the generated query information to the AI ​​server 40 and receives response information to the query information. The application processing unit 202 causes the output processing unit 204 to generate output information based on the received response information. The application processing unit 202 may also cause the evaluation unit 206 to evaluate the reliability of the received response information. The application processing unit 202 transmits the output information obtained from the output processing unit 204 to the terminal device 10.

[0029] The output processing unit 204 composes output information to be output from its own device based on the answer information input from the application processing unit 202. The output processing unit 204 outputs the composed output information to the application processing unit 202. The output processing unit 204 is configured to include a syntax analyzer (parser) that converts, for example, a data string indicating text, a number sequence, or the like that composes the answer information into human-visible output information represented as text, an image, or a combination thereof.

[0030] The response information received from the AI ​​server 40 may include reference information indicating the amount of power supply that was referenced when estimating an abnormality in the amount of power supply. When evaluating the reliability of the response information, the evaluation unit 206 compares the actual measured value of the amount of power supply that was included in the reference information and notified to the AI ​​server 40 with the amount of power supply indicated in the reference information. The evaluation unit 206 determines the reliability so that the higher the degree of agreement between the two, that is, the higher the degree of match.

[0031] The storage unit 208 stores various types of data. The data stored in the storage unit 208 includes data acquired by the application server 20 itself and data used in the operation of the application server 20. The storage unit 208 stores, for example, a processing history and a field inspection history. The processing history is history data formed by accumulating analysis information for one analysis process. The content of the analysis information for each analysis process will be described later. The on-site inspection history is history data that is a compilation of inspection information that indicates the status of inspections that were conducted for each abnormality that occurred in the amount of power supplied from the power source. The storage unit 208 may be configured to include a processing history database that stores the processing history, and an inspection history database that stores the on-site inspection history.

[0032] Next, an example of analysis processing according to this embodiment will be described. Fig. 4 is a flowchart illustrating the analysis processing according to this embodiment. In the following description, the functional configuration of each unit or device will also be mentioned. (Step S102) The application processing unit 202 waits for input information indicating a target period for analyzing the operating status of the power plant 50 from the terminal device 10. When receiving the input information, the application processing unit 202 identifies the target period indicated by the input information. (Step S104) The application processing unit 202 determines whether or not there is analysis information obtained by performing a past analysis on the specified target period in the processing history stored in the storage unit 208. If it is determined that there is analysis information (YES in step S104), the process proceeds to step S132. If it is determined that there is no analysis information (NO in step S104), the process proceeds to step S106.

[0033] (Step S106) The application processing unit 202 queries the database server 30 for power plants 50 whose actual values ​​for the amount of power supplied during the target period are sufficiently lower than the predicted values, as candidate power sources that are candidates for analysis. Here, the application processing unit 202 transmits candidate power source query information for querying the candidate power sources to the database server 30.

[0034] When the database server 30 receives candidate power source inquiry information from the application processing unit 202, it calculates the forecast-to-actual ratio for the target period based on the measured data and weather data for each power plant 50. The database server 30 references the measured data stored for each power plant 50 and reads, for example, the amount of power transmitted from the power plant 50 to the power grid as the actual value of the amount of power supplied during the target period indicated by the candidate power source inquiry information. The database server 30 also reads the weather during the target period for each power plant 50 from the weather data. The database server 30 calculates a first forecast value (sometimes referred to herein as a "physics model forecast value") of the amount of power generated using a known, predefined physical model based on the weather indicated by the weather data. For example, an indirect forecasting method can be used as the physical model. The indirect forecasting method calculates the amount of power generated as the product of the amount of solar radiation, the loss coefficient, and the system capacity. The amount of solar radiation depends on the date, time, and weather (including sunshine hours and cloud cover) as meteorological information at the installation site of the power plant 50. The loss coefficient depends on factors such as the installation location and installation angle of the power generator (e.g., solar panels) installed in the power plant 50, and objects blocking sunlight (e.g., other equipment, plants, etc.). The system capacity corresponds to the maximum amount of power generated per unit time by the power generator (i.e., the generated power). The database server 30 is preset with an estimation formula for estimating the amount of solar radiation from the weather during normal operation, the loss coefficient, and the system capacity.

[0035] The database server 30 may calculate a second predicted value (sometimes referred to herein as a "learning model predicted value") as a predicted value of the power generation amount using a machine learning model that has been trained in advance from the weather indicated by the weather data. As the machine learning model, for example, any of a neural network, a random forest, a support vector machine, etc. may be used. The machine learning model is trained in advance on the relationship between known weather and actual power generation amount. Through training, a training set including multiple data sets each including weather and actual power generation amount is used, and a parameter set of the machine learning model is determined so that the magnitude of the difference between the estimated value of power generation calculated from the weather using the machine learning model and the actual value is minimized for the entire training set.

[0036] The database server 30 determines one of the first predicted value and the second predicted value or the average value as the predicted value of the power generation amount of the power plant 50. The database server 30 may calculate the predicted value of the power generation amount for each frame as the predicted power generation amount, regardless of whether or not candidate power source inquiry information has been received from the application processing unit 202, and may store in advance prediction data indicating the calculated predicted power generation amount.

[0037] The database server 30 calculates the forecast-actual ratio as the ratio of the actual value to the predicted value of the amount of power supply determined for each power plant 50. When the actual value of the amount of power supply is equal to the predicted value, the ratio is 1. When the actual value is lower than the predicted value, the forecast-actual ratio is smaller than 1. The smaller the actual value is relative to the predicted value, the smaller the forecast-actual ratio. In other words, the forecast-actual ratio corresponds to an index that indicates the degree to which the actual value deviates from the predicted value.

[0038] (Step S108) The database server 30 identifies as candidate power sources those power plants 50 whose forecast-to-actual ratio is equal to or less than a predetermined ratio (for example, 0.4 to 0.8, typically 0.5) that is less than 1, and creates a power plant list showing the identified candidate power sources. The database server 30 sends the created power plant list to the application processing unit 202 as a response to the candidate power source inquiry information. The application processing unit 202 acquires a list of power plants indicating candidate power sources from the database server 30 .

[0039] If the target period includes multiple time slots, the database server 30 may determine predicted and actual values ​​for the amount of power supply for each time slot and calculate the forecast-to-actual ratio for each time slot based on these values. In this case, the database server 30 may identify as a candidate power source a power plant 50 whose forecast-to-actual ratio is equal to or less than a predetermined ratio for even a portion of the time slots (e.g., one time slot) within the target period. The database server 30 may determine the average values ​​of the predicted and actual values ​​for each time slot within the target period, determine one set of forecast-to-actual ratios based on these values, and identify candidate power sources based on the forecast-to-actual ratio. If there is no power plant 50 whose forecast-to-actual ratio is equal to or less than the predetermined ratio, the database server 30 transmits response information to the application processing unit 202 indicating that there is no candidate power source.

[0040] (Step S110) For each unprocessed candidate power source, the application processing unit 202 performs the processes of steps S112 to S130, treating that candidate power source as the processing target (step S110 YES). If there are no unprocessed candidate power sources (step S110 NO), the process proceeds to step S132. (Step S112) The application processing unit 202 requests the database server 30 to provide the actual measurement data, forecast data, and weather data for the target period of the power plant 50 to be processed, and acquires this data from the database server 30. (Step S114) The application processing unit 202 reads the field inspection history from the storage unit 208. (Step S116) The application processing unit 202 stores a series of acquired information related to the candidate power source, i.e., data including actual measurement data, forecast data, weather data, and field inspection history, as part of new analysis information in the storage unit 208. The stored acquired information is accumulated as part of the processing history.

[0041] (Step S120) The application processing unit 202 constructs a prompt that indicates query information for inquiring about the cause of an abnormality in the amount of power supplied from the candidate power source, including actual measurement data and forecast data related to the candidate power source. The application processing unit 202 may further include a site inspection history in the query information. In other words, by sending a prompt indicating query information including the site inspection history to the AI ​​server 40, the information processing system S1 can also be considered a retrieval augmented generation (RAG) system that generates answer information based on the query information. Here, the application processing unit 202 parses the constructed prompt to break it down into multiple pieces of information, and then uses a predetermined conversion method to transmit a data string indicating each piece of information to the AI ​​server 40.

[0042] The AI ​​server 40 receives a data string representing query information from the application processing unit 202. The AI ​​server 40 converts the received data string into an input token string, which is a permutation including multiple tokens (tokenization). Each token constitutes a processing unit in the AI ​​model. The AI ​​server 40 uses a trained AI model to calculate an output token string representing response information from the input token string representing query information. During the calculation process, the AI ​​server 40 estimates the cause of an abnormality in the performance values ​​indicated by the actual measurement data and predicted data included in the query information and the degree of necessity for an on-site inspection based on a portion of the inspection information included in the on-site inspection history. For example, the AI ​​server 40 may refer to past inspection information based on the actual measurement data and predicted data that have a correlation that results in similar performance values ​​due to the statistical properties of the abnormality, with more emphasis than other inspection information. When estimating the cause of the abnormality and the degree of necessity for an on-site inspection, the AI ​​server 40 may refer to actual measurement data and predicted data that are separate from the actual measurement data and predicted data included in the query information. The AI ​​server 40 may refer to separate actual measurement data and predicted data, such as data included in past reference information and notified, data added to past investigation information, or fictitious predicted data and actual measurement data estimated from these data. The AI ​​server 40 may include the actual measurement data and predicted data referenced when estimating the cause of the abnormality and the degree of need for an on-site investigation as reference information in the response information. The AI ​​server 40 converts the output token sequence representing the response information into an output data sequence having a predetermined format (detokenization). The output data sequence may be, for example, text, a number sequence, or the like, and transmits it to the application processing unit 202. Note that tokenization, detokenization, or both may be performed by the application processing unit 202 instead of the AI ​​server 40.

[0043] (Step S122) The application processing unit 202 receives an output data string indicating the answer information from the AI ​​server 40, and outputs the output data string indicating the received answer information to the output processing unit 204, which then creates output information based on the answer information. The output processing unit 204 performs syntax analysis on the output data string indicating the answer information and converts it into text and images indicating the answer information. The output processing unit 204 generates a display screen that forms the output information by arranging the converted text and images to have a predetermined structure. As the output information, for example, a display screen is generated that shows the cause of an abnormality in the actual value of power supplied from the candidate power source indicated in the answer information and the degree of necessity for an on-site investigation as analysis results. The output processing unit 204 outputs the constructed output information to the application processing unit 202. (Step S124) The application processing unit 202 stores data including the query information sent to the AI ​​server 40, the response information received from the AI ​​server, and the output information input from the output processing unit 204 as another part of the new analysis information in the memory unit 208. The stored data is accumulated as part of the processing history.

[0044] (Step S126) When the answer information received from the AI ​​server 40 includes reference information, the application processing unit 202 outputs the received answer information to the evaluation unit 206 and requests an evaluation of the reliability of the answer information. (Step S128) The evaluation unit 206 compares the actual measurement data and predicted data indicated in the reference information with the actual measurement data and predicted data analyzed by the AI ​​server 40, and evaluates the reliability of the response information based on the degree of match between the two. More specifically, the evaluation unit 206 reads the predicted data and actual measurement data for the target period of the candidate power source to be processed from the storage unit 208. The evaluation unit 206 compares the second predicted value indicated in the predicted data specified in the reference information and the second actual value indicated in the actual measurement data with the first predicted value indicated in the read predicted data and the first actual value indicated in the actual measurement data, respectively. The evaluation unit 206 determines whether the second predicted value matches the first predicted value, and whether the second actual value matches the first actual value. The evaluation unit 206 calculates the ratio of the number of samples of the second actual value to the number of samples of the first actual value among the number of samples of the second predicted value and the first predicted value to be evaluated, and determines the reliability based on the calculated ratio. The evaluation unit 206 may, for example, determine the calculated ratio as the reliability. The evaluation unit 206 may set a code indicating the reliability for each of a plurality of predetermined value ranges of ratios, identify which of the predetermined value ranges the calculated ratio belongs to, and determine the code corresponding to the identified value range as the reliability. The evaluation unit 206 outputs the determined reliability to the application processing unit 202.

[0045] (Step S130) The application processing unit 202 stores the reference information output to the evaluation unit 206 and the data indicating the reliability input from the evaluation unit 206 in the storage unit 208 as further parts of the new analysis information. The stored data is accumulated as part of the processing history. Note that the application processing unit 202 may store the reliability notified from the evaluation unit 206 in association with the candidate power source targeted for processing and the output information related to the target period. Then, the process returns to step S110.

[0046] (Step S132) The application processing unit 202 reads the output information included in the analysis information related to the target period from the processing history stored in the storage unit 208. The application processing unit 202 transmits the read output information to the terminal device 10, and causes it to be presented to the user. The user interface unit 102 of the terminal device 10 outputs, as the output information received from the application processing unit 202, for example, display data showing a display screen to the display 114.

[0047] Next, a specific example of data to be processed in this embodiment will be described. FIG. 5 is a diagram illustrating actual measurement data and predicted data according to this embodiment. FIG. 5 illustrates time series of actual and predicted values ​​of the amount of power supplied from a power plant 50 having a solar power generation system. In FIG. 5, the vertical axis represents the amount of power generated, and the horizontal axis represents date and time. The physical model predicted value is a predicted value of the amount of power generated calculated from weather using a physical model. The TSO (Transmission System Operator) actual value is an actual value of the amount of power transmitted from the power plant 50 to the power grid of the power transmission operator. In other words, the TSO actual value corresponds to an actual measured value of the amount of power transmitted. The PCS (Power Conditioning System) actual value is an actual measured value of the amount of AC power obtained by converting DC power generated by solar panels. In other words, the PCS actual value corresponds to an actual measured value of the amount of power generated. The TSO actual value and the PCS actual value constitute actual measurement data. The physical model predicted value constitutes prediction data.

[0048] The application processing unit 202 of the application server 20 may calculate a total of four forecast-to-actual ratios by using the TSO actual value and the PCS actual value as actual measured values ​​of supply volume and the physical model predicted value and the learning model predicted value as predicted values ​​of supply volume. In this case, the application processing unit 202 compares each of the four forecast-to-actual ratios with a predetermined ratio and identifies, as a candidate power source, a power plant 50 for which any one of the forecast-to-actual ratios is equal to or less than the predetermined ratio. Alternatively, the application processing unit 202 may calculate a total of two forecast-to-actual ratios by using the TSO actual value and the PCS actual value as actual measured values ​​of supply volume and the average of the physical model predicted value and the learning model predicted value as a predicted value of supply volume. In this case, the application processing unit 202 compares each of the two forecast-to-actual ratios with a predetermined ratio and identifies, as a candidate power source, a power plant 50 for which any one of the forecast-to-actual ratios is equal to or less than the predetermined ratio.

[0049] In the example of Figure 5, the physics model prediction value, TSO actual value, and PCS actual value all fluctuate on a daily cycle. The TSO actual value and the PCS actual value are nearly identical. In contrast, the physics model prediction value does not necessarily match the TSO actual value. Up until 20YY / MM / 02, the TSO actual value is close to the physics model prediction value, but from 20YY / MM / 03 onward, the TSO actual value is significantly lower than the physics model prediction value. Therefore, the power plant 50 shown in Figure 5 can be a candidate power source. Furthermore, from Figure 5, it is inferred that a cause occurred that caused the TSO actual value to decrease between 20YY / MM / 02 and 20YY / MM / 03.

[0050] FIG. 6 illustrates measured data, forecast data, and weather data according to this embodiment. As illustrated in FIG. 6, the measured data, forecast data, and weather data may be integrated into a single data file. In the example of FIG. 6, each row contains the date and time for each frame, and each column contains the TSO actual value, PCS actual value, learning model forecast value, physics model forecast value, latest forecast value, temperature, global solar radiation, previous time global solar radiation, previous time precipitation, previous time snowfall, wind direction, wind speed, dew point temperature, and relative humidity, which are associated with each other. The temperature, global solar radiation, previous time global solar radiation, previous time precipitation, previous time snowfall, wind direction, wind speed, dew point temperature, and relative humidity are meteorological element information constituting the weather data. Some of this element information is used to predict power generation. The previous time global solar radiation, previous time precipitation, and previous time snowfall respectively indicate the global solar radiation, precipitation, and snowfall at the time corresponding to the immediately preceding frame. The data file illustrated in FIG. 6 is, for example, a data file in CSV format.

[0051] FIG. 7 is a diagram illustrating a power plant list according to this embodiment. The power plant list includes a power plant ID, a power plant name, and an installed capacity for each power plant 50 managed by the user (business operator), and is configured by associating these. A list of all power plants 50 that are the subject of management is stored in the database server 30. In the processing of step S108, a power plant list is provided that indicates the power plants 50 that are candidate power sources and that will become part of the list.

[0052] 8 is a diagram showing a first example of investigation information according to this embodiment. The investigation information is information showing the results of an investigation into an abnormality in the amount of power supply, that is, an event in which the actual amount of power generation or transmission is significantly lower than its predicted value. The investigation information illustrated in FIG. 8 indicates the power plant ID of the power plant 50 to be investigated, the power plant name, the event that occurred in the amount of power generated, and the investigation date and time. In the example of Figure 8, the event is described as both the actual measured value of power generation (PCS actual value) and the actual measured value of power transmission (TSO actual value) being lower than the predicted value of power supply (power generation). The cause is listed as shielding by surrounding trees, and the countermeasure is listed as tree cutting. The date and time of the countermeasure is the date and time during or immediately after the implementation of the countermeasure. If the cause is unknown, this is also stated. In that case, the countermeasure is not implemented, and so the relevant description may not be listed in the columns for countermeasure and countermeasure date and time.

[0053] FIG. 9 is a diagram showing a second example of investigation information according to this embodiment. In the example of FIG. 9, an event occurring in the amount of power supply is described in which the actual measured value of the amount of power transmitted (TSO actual value) is lower than the predicted value of the amount of power supplied (amount of power generated). A power transmission stop command from the grid (power grid) is listed as the cause. When the amount of power generated significantly exceeds the amount of power demand, the power grid may issue a power transmission stop command to the power source targeted for power transmission stop in order to limit the amount of power supplied from the power source. The countermeasure column describes leaving the situation as it is and waiting until the power transmission stop command is lifted. In this example, there is no significant difference between the actual measured value of the amount of power transmitted and the predicted value of the amount of power generated, and the discrepancy between the actual measured value of the amount of power transmitted and the predicted value of the amount of power generated is significant. Therefore, it is inferred that there is little possibility of a malfunction in the solar power generation system, and there is a possibility of a malfunction related to power transmission to the power grid or a malfunction related to the node. The investigation information may include actual measurement data indicating actual values ​​of events occurring in the power supply, forecast data indicating forecast values, or reference information indicating the location of such data (e.g., URL (uniform resource locator), file path, address, etc.). The prompt may also include a field investigation history containing the investigation information. The AI ​​server 40 can refer to the actual measurement data and forecast data indicating events occurring in the power supply and analyze the correlation between the events and the actual measurement data and forecast data. This allows the AI ​​server 40 to more accurately determine the cause of the event and the need for investigation.

[0054] FIG. 10 is a diagram showing an example of a schematic configuration of a prompt according to this embodiment. The prompt represents query information for inquiring the AI ​​server 40 about the cause of an abnormality in the actual measured value of the power supplied from the power plant 50 during the target period. An abnormality in the actual measured value is detected by comparing it with a predicted value predicted from the weather. The prompt illustrated in FIG. 10 includes actual measured data, predicted data, weather data, and on-site inspection history. The actual measured data indicates the PCS actual value and the TSO actual value for each frame during the target period. The predicted data indicates the learning model predicted value and the physical model predicted value for each frame during the target period. The weather data indicates the weather at the installation location of the power plant 50 for each frame during the target period. The on-site inspection history is composed of the accumulation of inspection information for each on-site inspection. The on-site inspection history illustrated in FIG. 10 includes M pieces of inspection information 01 to M (M is an integer greater than or equal to 1).

[0055] FIG. 11 is a diagram illustrating a specific example of a prompt according to this embodiment. The prompt illustrated in FIG. 11 describes the role, the request, power plant data, and past investigation information. The role describes the functions that the AI ​​server 40 should have. More specifically, the content of the data analysis related to a solar power plant includes analyzing the cause of a large discrepancy between the predicted power generation amount and the transmitted power amount. Possible causes include an inaccurate weather forecast, communication errors and improper maintenance between the power regulation system and the power plant, adhesion of debris on the power generation panels, shielding of the power generation panels, supply control from the power grid, and failure of measuring equipment. The request describes the function to be executed by the AI ​​server 40. More specifically, for anomalies indicated in the power plant data for the target period, an analysis report in a predetermined format is created by referring to the on-site investigation history. Response information from the AI ​​server 40 is written in the analysis report. The response information indicates the cause of the anomaly and the need for an on-site investigation. Note that the power plant data and past investigation information are attached to the prompt as reference information when executing the function specified in the request. Power plant data includes actual measurement data, forecast data, and weather data. csv {actual_data} indicates a description field into which power plant data named actual_data in csv format is inserted. Past survey information corresponds to field survey history. csv {research_data} indicates a description field into which field survey history named research_data in csv format is inserted.

[0056] FIG. 12 is a diagram showing a first example of response information according to this embodiment. The response information has items for priority, reason for priority, and summary. As priority, a value of 1 out of 3 is entered as a number indicating the degree of necessity for an on-site investigation. A higher number indicates a lower necessity for an on-site investigation. As reasons for priority, the necessity for an on-site investigation is described based on the presence or absence of similar events included in past investigation history, possible causes, and the degree of impact of those causes. In the example of FIG. 12, it is stated that an equipment failure that has not occurred in the past is assumed to be the cause. As a summary, the cause estimated based on the decrease in power generation shown in the power plant data and the necessity for that cause are described in more detail.

[0057] FIG. 13 is a diagram showing a second example of response information according to this embodiment. In the example of FIG. 13, a value of 3 is entered as the numerical value indicating the priority. The reason for the priority is given as the possibility that voltage suppression or output suppression is being performed as a known phenomenon that cannot be resolved even by an on-site investigation. In summary, the cause estimated based on the reduction in power transmission volume shown in the power plant data and the necessity for that cause are described in more detail.

[0058] In this embodiment, as illustrated in Fig. 14, each time new investigation information is acquired through an on-site investigation, that investigation information is added and accumulated as an on-site investigation history. By including the on-site investigation history in the prompt indicating the inquiry information, an event that actually occurred as the cause of the abnormality can be more accurately estimated. The process in Fig. 14 includes steps S202 to S208. (Step S202) The user interface unit 102 of the terminal device 10 acquires new survey information in response to a user operation. (Step S204) The user interface unit 102 uploads the acquired survey information to the application server 20. (Step S206) The application processing unit 202 of the application server 20 stores the survey information received from the terminal device 10 in the storage unit 208. (Step S208) The storage unit 208 updates the field investigation history by adding the investigation information newly provided from the application processing unit 202. The storage unit 208 notifies the application processing unit 202 of an update completion response.

[0059] 12 and 13, the AI ​​server 40 derives the answer information by referring to the on-site investigation history described in the prompt. Therefore, the application processing unit 202 configures the prompt to include the on-site investigation history, which is updated by adding new investigation information each time an on-site investigation is conducted, so that the information on the cause and countermeasures indicated in the investigation information is referenced when estimating the cause of an abnormality in the amount of power supplied from the power plant 50 and the need for an on-site investigation. Therefore, by including in the prompt an on-site investigation history with sufficient accumulated investigation information, the cause of the abnormality and the need for an on-site investigation can be estimated as accurately as or even more accurately than a skilled worker.

[0060] Fig. 15 is a diagram showing a first example of a display screen according to this embodiment. The display screen shown in Fig. 15 has items for specifying an analysis period and for locations where the actual value / physical model predicted value is 50% or less. The analysis period is specified by a user operation in an input field for specifying the analysis period. A site where the actual value / physical model predicted value is 50% or less refers to a candidate power source where the ratio of actual to forecast is 50% or less. A list of power plants showing the power plants 50 that are candidate power sources is displayed here. The "Perform detailed analysis" button is a button that, when pressed, commands the execution of analysis processing. Directly below the "Perform detailed analysis" button, an overview of the output information that constitutes the analysis results is listed in each row for each candidate power source. By specifying one of the rows through user operation, a command is given to display more detailed output information for the candidate power source corresponding to that row.

[0061] FIG. 16 is a diagram showing a second example of a display screen according to this embodiment. The display screen shown in FIG. 16 shows output information for SOLAR01, a candidate power source, in processing details. The left side of FIG. 16 shows, in that order, the physical model prediction value by day, the TSO actual value, the PCS actual value, and the physical model prediction value by frame, the TSO actual value, and the PCS actual value. The right side of FIG. 16 shows an overview of the entire period, dates and items that should be checked in particular, and the order of priority for each item. This information is extracted from the response information received from the AI ​​server 40.

[0062] FIG. 17 is a sequence chart illustrating the analysis process according to this embodiment. (Step S302) The application processing unit 202 of the application server 20 identifies the target period indicated by the input information received from the terminal device 10. (Step S304) The application processing unit 202 queries the storage unit 208 to see whether there is analysis information based on the power plant data within the target period. (Step S306) The application processing unit 202 acquires from the storage unit 208 a response regarding the presence or absence of analysis information.

[0063] (Step S308) When the application processing unit 202 receives a response indicating that analysis information is available, it extracts output information from the analysis information related to the target period. The application processing unit 202 transmits the extracted output information to the terminal device 10 to have it presented. Thereafter, the processing in FIG. 17 ends. When the application processing unit 202 receives a response indicating that there is no analysis information, the process proceeds to step S310 and subsequent steps.

[0064] (Step S310) The application processing unit 202 queries the database server 30 for the forecast / actual ratio for each power plant 50 for the target period. The database server 30 calculates the forecast / actual ratio from the predicted value of the amount of power supply calculated based on the weather data for the target period and the actual value of the amount of power supply indicated in the actual measurement data. (Step S312) The application processing unit 202 receives a response indicating the forecast / actual ratio for each power plant 50 from the database server 30. (Step S314) The application processing unit 202 queries the database server 30 for power plants as candidate power sources whose budget-to-actual ratios are equal to or less than a predetermined ratio. The database server 30 identifies power plants 50 whose calculated budget-to-actual ratios are equal to or less than the predetermined ratio, and creates a power plant list showing the identified power plants 50. (Step S316) The application processing unit 202 receives a power plant list indicating the identified power plant 50 from the database server 30.

[0065] The application processing unit 202 executes the processing of steps S318 to S334 for each power plant 50 shown in the power plant list, and then proceeds to the processing of step S336. (Step S318) The application processing unit 202 requests the database server 30 for actual measurement data related to the candidate power source for the target period. (Step S320) The application processing unit 202 receives the requested actual measurement data from the database server 30. (Step S322) The application processing unit 202 requests the database server 30 for forecast data related to the candidate power sources for the target period. (Step S324) The application processing unit 202 receives the requested prediction data from the database server 30. (Step S326) The application processing unit 202 requests the database server 30 for meteorological data relating to the installation locations of the candidate power sources for the target period. (Step S328) The application processing unit 202 receives the requested weather data from the database server 30.

[0066] (Step S330) The application processing unit 202 requests the field inspection history stored in the storage unit 208. (Step S332) The application processing unit 202 acquires the field inspection history from the storage unit 208. (Step S334) The application processing unit 202 stores the acquired actual measurement data, forecast data, and meteorological data together with the estimated-actual ratio for the power plant 50 being processed in the storage unit 208 as part of the analysis information.

[0067] The application processing unit 202 executes the processing of steps S336 to S356 for each power plant 50 shown in the power plant list, and then proceeds to the processing of step S358. (Step S336) The application processing unit 202 composes a prompt indicating query information including the acquired actual measurement data, forecast data, weather data, and on-site inspection history, and sends the composed prompt to the AI ​​server 40. Using a pre-trained analytical model, the AI ​​server 40 estimates, as analysis results, the cause of the abnormality that has occurred in the actual measurement value of the power supply and the need for an investigation from the query information indicated in the prompt received from the application processing unit 202. (Step S338) The application processing unit 202 receives answer information indicating the estimated analysis result from the AI ​​server 40.

[0068] (Step S340) The application processing unit 202 outputs the acquired answer information to the output processing unit 204, and causes the output processing unit 204 to compose output information based on the answer information. (Step S342) The application processing unit 202 receives output information from the output processing unit 204. (Step S344) The application processing unit 202 stores the inquiry information, response information, and output information acquired for the power plant 50 to be processed in the storage unit 208 as another part of the analysis information.

[0069] (Step S346) The application processing unit 202 requests the evaluation unit 206 to evaluate the reliability of the answer information acquired from the AI ​​server 40. (Step S348) The evaluation unit 206 requests the storage unit 208 for the actual measurement data and predicted data for the target power plant 50 for the target period. (Step S350) The evaluation unit 206 receives the requested actual measurement data and predicted data. (Step S352) The evaluation unit 206 compares the actual measurement data and the predicted data indicated by the reference information included in the response information, and calculates the reliability based on the degree of agreement between the two. (Step S354) The application processing unit 202 receives a response of the reliability from the evaluation unit 206. (Step S356) The application processing unit 202 stores the reference information and the reliability in the storage unit 208 as another part of the analysis information.

[0070] (Step S358) The application processing unit 202 transmits the output information for each power plant 50 that is a candidate power source to the terminal device 10, causing it to be presented. The application processing unit 202 may cause the output information derived from the response information that is the evaluation target of reliability to be included in the output information and presented. Thereafter, the processing of FIG. 17 ends.

[0071] The terminal device 10, application server 20, database server 30, and AI server 40 can each be configured as an information processing device equipped with a computer system. Figure 18 is a schematic block diagram showing an example of the configuration of a computer system 60 according to this embodiment.

[0072] The computer system 60 includes, for example, a processor 62, a storage 64, an input / output I / F (Interface) 66, a communication I / F 68, a ROM 72, and a RAM 74. The processor 62, the storage 64, the input / output I / F 66, the communication I / F 68, the ROM 72, and the RAM 74 are connected to each other via a bus BS so that various types of data can be input and output to and from each other.

[0073] The processor 62 reads out programs and various data stored in the ROM 72, executes the programs, and controls the operation of the computer system 60 and devices that include the computer system 60. For example, the processors of the terminal device 10 and the application server 20 execute predetermined application programs to realize the functions of the user interface unit 102 and the application processing unit 202, respectively. In this application, "executing a program" or "executing a program" includes the meaning of executing processing instructed by instructions written in the program. The processor 62 is, for example, a CPU (Central Processing Unit). The processor 62 may include a GPU (Graphics Processing Unit) in addition to the CPU.

[0074] The storage 64 is an auxiliary storage device that can continuously store various types of data in a readable and writable manner. The storage 64 may be, for example, a hard-disk drive (HDD) or a solid-state drive (SSD). The input / output I / F 66 connects to other devices via wire or wirelessly in accordance with a predetermined input / output standard so as to be able to input and output various types of data. The communication I / F 68 connects to other devices via wire or wirelessly in accordance with a predetermined communication standard so as to be able to input and output various types of data.

[0075] The ROM 72 stores, for example, programs to be executed by the processor 62 or various devices. The RAM 74 is used as a main storage medium that functions as a work area for temporarily storing various data and programs used by the processor 62, for example.

[0076] In the above description, the individual power sources are mainly solar power plants equipped with solar power generation systems, but this is not limiting. Instead of or in addition to solar power plants, this embodiment may also be applied to other types of power sources, such as wind power plants and hydroelectric power plants, whose power supply depends on the weather. In the above description, the database server 30 calculates the forecast / actual ratio for each power plant 50, but this is not limiting. Instead of the database server 30, any functional unit of the application server 20 (for example, the application processing unit 202) may perform the calculation. Although the predicted-actual ratio is mainly used as an index indicating the degree of deviation between the predicted value and the actual measured value of the amount of power supply, this is not limiting. In this embodiment, other types of indexes, for example, a normalized residual, may be used instead of or in addition to the predicted-actual ratio. The normalized residual is an index obtained by dividing the difference between the actual measured value and the predicted value by a reference value. Either the predicted value or the standard deviation of the actual measured value may be used as the reference value. By definition, an index may be applied in which the larger the value, the greater the deviation between the predicted value and the actual measured value.

[0077] In the above description, the terminal device 10, application server 20, database server 30, and AI server 40 are each an independent device, but this is not limiting. Any combination of the components of these devices may be implemented as a single device. Furthermore, some components or functions may be omitted, or other components or functions may be added to any of the devices.

[0078] For example, the application server 20 may be configured as a single device having the configuration of the database server 30. The terminal device 10 may be configured as a single device having the configuration of the application server 20 or both the application server 20 and the database server 30. The application server 20 may be configured as a single device having the configuration of the AI ​​server 40 or both the application server 20 and the database server 30.

[0079] As described above, the information processing device (e.g., application server 20) according to this embodiment acquires actual measurement data indicating the actual measurement value of the amount of electricity supplied for each power source (e.g., power plant 50), selects as candidate power sources to be investigated those for which the deviation between the predicted value of the amount of electricity supplied, calculated using a predetermined mathematical model based on weather data indicating the weather at the installation location of the power source, and the actual measurement value is equal to or greater than a predetermined reference value (e.g., the predicted-to-actual ratio is equal to or less than a predetermined ratio), generates query information (e.g., a prompt) for inquiring about the cause of the abnormality in the amount of electricity supplied, including the actual measurement value (e.g., TSO actual value and PCS actual value indicated by the actual measurement data) and the predicted value (e.g., learning model predicted value and physical model predicted value of the amount of electricity generated indicated by the predicted data) of the candidate power source, outputs the query information to a predetermined analysis model (e.g., AI server 40), and acquires response information from the analysis model including the cause of the abnormality and the need for an investigation (e.g., an on-site investigation). This configuration narrows down the analysis targets by selecting as candidate power sources those power sources where the actual measured power supply volume deviates significantly from the forecasted value based on weather. Based on the actual measured and forecasted values ​​for the candidate power sources, the analytical model estimates response information indicating the cause of the abnormal power supply volume and the need for investigation. This improves the efficiency of operational management of multiple power sources.

[0080] Furthermore, the information processing device according to this embodiment may configure the inquiry information by further including investigation information (for example, on-site investigation history) relating to investigations of past abnormalities. According to this configuration, by further referring to investigation information relating to past investigations, response information that more accurately indicates the cause of the abnormality and the need for investigation can be obtained.

[0081] In this embodiment, the response information may include reference information (e.g., actual measurement data, predicted data) indicating the amount of power supply referenced when estimating the cause of the abnormality in the amount of power supply. The information processing device may compare the actual measurement value and predicted value of the amount of power supply from the power source with the amount of power supply indicated in the reference information to evaluate the reliability of the response information. According to this configuration, the reliability of the answer information is evaluated based on whether the actual measured values ​​and predicted values ​​actually provided to the analysis model were referenced when estimating the answer information by comparing the reference information referred to when estimating the answer information with the actual measured values ​​and predicted values ​​included in the inquiry information as the analysis target. Therefore, it is possible to reduce or eliminate the risk of obtaining answer information by referencing fictitious power supply (hallucination risk).

[0082] In addition, in this embodiment, the ratio of the actual measured value to the predicted value (i.e., the predicted-actual ratio) may be calculated as an index value of the magnitude of the deviation, and power sources for which this ratio is equal to or less than a predetermined reference ratio that is smaller than 1 may be selected as candidate power sources. With this configuration, power supplies whose predicted values ​​of power supply are lower than the actual measured values ​​are quantitatively selected as analysis targets.

[0083] The power source (for example, the power plant 50) may include a solar power generation system. The weather data may include information on the amount of solar radiation at the installation location of the power source for each unit time (for example, each frame).

[0084] The mathematical model may include a physical model for calculating a first predicted value of the amount of electricity supply from the amount of solar radiation at the installation location of the power source, and a trained machine learning model for calculating a second predicted value of the amount of electricity supply from weather data, and the query information may include the first predicted value and the second predicted value.

[0085] In addition, the information processing system S1 according to this embodiment may include the above-mentioned information processing device and a model processing unit (e.g., AI server 40) that uses an analytical model to estimate the cause of an abnormality and the need for investigation of the abnormality from inquiry information, and outputs response information including the estimated cause and the need for investigation to the information processing device.

[0086] Furthermore, an information processing method in the information processing device according to this embodiment may involve the information processing device acquiring actual measurement data indicating the actual measured value of the amount of power supplied for each power source, selecting as candidate power sources to be investigated power sources for which the degree of deviation indicating the magnitude of deviation between the predicted value of the amount of power supplied, calculated using a predetermined mathematical model based on weather data indicating the weather at the installation location of the power source, and the actual measured value is equal to or greater than a predetermined reference value, generating inquiry information for inquiring about the cause of any abnormality in the amount of supply, including the actual measured value and the predicted value of the candidate power source, outputting the inquiry information to a predetermined analysis model, and acquiring response information from the analysis model including the cause of the abnormality and the need for investigation.

[0087] Although the embodiments of the present invention have been described above in detail with reference to the drawings, the specific configurations are not limited to the above-described embodiments, and the present invention also includes designs that do not deviate from the gist of the present invention. The configurations described in the above-described embodiments can be combined in any manner. [Explanation of symbols]

[0088] S1...information processing system, 10...terminal device, 20...application server, 30...database server, 40...AI server, 50 (50-1 to 50-N)...power plant, 60...computer system, 62...processor, 64...storage, 66...input / output I / F, 68...communication I / F, 72...ROM, 74...RAM, 102...user interface unit, 112...input device, 114...display, 202...application processing unit, 204...output processing unit, 206...evaluation unit, 208...storage unit

Claims

1. Obtain actual measurement data showing the actual amount of power supplied by each power source, selects, as candidate power sources to be investigated, power sources for which the magnitude of deviation between the predicted value of the amount of power supply calculated using a predetermined mathematical model based on meteorological data indicating the weather at the installation location of each power source and the actual measured value is equal to or greater than a predetermined reference value; generating inquiry information for inquiring about the cause of the abnormality in the supply amount, including the actual measured value and the predicted value of the candidate power source; outputting the query information to a predetermined analytical model; Obtaining response information from the analytical model, including the cause of the anomaly and the need for investigation. Information processing device.

2. The information processing apparatus according to claim 1 , wherein the query information further includes investigation information relating to investigations into the abnormality in the past.

3. the response information includes reference information indicating the amount of power supply referenced when the cause is estimated, The actual measured value and the predicted value are compared with the supply amount indicated in the reference information to evaluate the reliability of the response information. The information processing device according to claim 1 .

4. calculating a ratio of the actual measurement value to the predicted value as an index of the magnitude of the deviation; The power source having the ratio equal to or less than a predetermined reference ratio is selected as the candidate power source. The information processing device according to claim 1 .

5. the power source comprises a solar power generation system; The meteorological data includes at least information on the amount of solar radiation per unit time at the installation location of the power source. The information processing device according to claim 1 .

6. the mathematical model includes a physical model for calculating a first predicted value of the supply amount from the amount of solar radiation, and a trained machine learning model for calculating a second predicted value of the supply amount from the meteorological data, The query information includes the first predicted value and the second predicted value. The information processing device according to claim 5 .

7. The information processing device according to claim 1 ; using the analytical model to estimate a cause of the anomaly and the need for an investigation of the anomaly from the inquiry information; a model processing unit that outputs answer information including the cause and the necessity to the information processing device. Information processing system.

8. An information processing method in an information processing device, Obtain actual measurement data showing the actual amount of power supplied by each power source, selects, as candidate power sources to be investigated, power sources for which the magnitude of deviation between the predicted value of the amount of power supply calculated using a predetermined mathematical model based on meteorological data indicating the weather at the installation location of each power source and the actual measured value is equal to or greater than a predetermined reference value; generating inquiry information for inquiring about the cause of the abnormality in the supply amount, including the actual measured value and the predicted value of the candidate power source; outputting the query information to a predetermined analytical model; Obtaining response information from the analytical model, including the cause of the anomaly and the need for investigation. Information processing methods.

Images