Event forecasting support system and event forecasting support method

Abstract

To efficiently and quickly create a scenario for forecast of an event such as weather.SOLUTION: An event forecast support system generates an event forecast error matrix based on event forecast result information and event result information, generates an average value vector having the average values of forecast targets representing an event as elements and a standard deviation vector having a standard deviation as an element, which are determined based on the event forecast error matrix, generates a correlation matrix between the forecast targets based on the event forecast error matrix, determines a characteristic value and a characteristic vector by performing characteristic value analysis on the correlation matrix, specifies the characteristic value and the characteristic vector of a main component, generates, by using random numbers, a plurality of forecast error vectors that are a vector having errors in the forecast targets as elements based on the average value vector, standard deviation vector, and specified characteristic value and characteristic vector, and reflects the plurality of generated forecast error vectors on an event forecast in a forecast period acquired from the event forecast result information to create a scenario for the plurality of event forecasts.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to an event prediction support system and an event prediction support method.

Background Art

[0002] Conventionally, various mechanisms have been proposed to obtain weather information that meets the needs of the user by verifying and correcting errors in the weather information provided by the Japan Meteorological Agency or the like.

[0003] For example, Patent Document 1 describes a weather prediction error analysis system configured for the purpose of enabling the accuracy of weather prediction information to be determined. The weather prediction error analysis system calculates a predicted value of a physical quantity related to a specific weather phenomenon by applying weather observation data to a weather model, and stores a conditional probability table determined based on the dependence relationship with a plurality of weather determination elements for each error category determined from the difference between the predicted value and the weather observation data. Based on the determination result of the weather determination element, a probability distribution related to the error category is calculated from the conditional probability table.

[0004] For example, Patent Document 2 describes a power demand forecasting processing device configured for the purpose of appropriately forecasting demand or power generation based on ensemble forecasts. The demand forecasting processing device stores actual power demand values ​​and weather forecast values ​​for multiple forecast times at one or more locations for each distribution day for multiple scenarios, extracts weather forecast values ​​for multiple scenarios at a specified location, and extracts actual power demand values ​​at a virtual demand forecast date and time obtained by adding a predetermined distribution time and forecast interval for each day within the distribution period to be processed. The demand forecasting processing device then calculates coefficients for a demand forecasting model having a term that adjusts the height of the demand forecast value distribution by the median or mean of the weather forecast values ​​for multiple scenarios, and a term that adjusts the width of the demand forecast value distribution by the difference between the weather forecast value and the median or mean of the weather forecast value, so that the distribution of demand forecast values ​​calculated by the weather forecast values ​​and the demand forecasting model follows user settings based on actual power demand values, applies weather forecast values ​​for multiple scenarios at a specified location, with the forecast interval on the specified distribution day as the forecast time, to the demand forecasting model with the coefficients set for the demand forecasting model, and calculates multiple sets of demand forecast values. [Prior art documents] [Patent Documents]

[0005] [Patent Document 1] Japanese Patent Publication No. 2018-10015 [Patent Document 2] Japanese Patent Publication No. 2009-225550 [Overview of the Initiative] [Problems that the invention aims to solve]

[0006] As described in Patent Document 2, one method for improving the accuracy of weather forecasts is to use multiple independently generated weather forecast scenarios (ensemble method). However, the weather forecast history information and weather performance information used to generate the scenarios contain a large amount of information, and if there are many forecast areas and forecast dates and times, it is necessary to handle an enormous amount of information. When attempting to build a practical system for automatically generating weather forecast scenarios using such a method, there were challenges in terms of efficient use of information processing resources and speed.

[0007] This invention was made in view of the above background, and aims to provide an event forecasting support system and an event forecasting support method that can efficiently and quickly generate forecasting scenarios for weather and other events. [Means for solving the problem]

[0008] One of the present inventions for achieving the above objective is an event forecasting support system, configured using an information processing device having a processor and memory, which stores a forecast period, which is the period for which an event is forecasted, event forecasting performance information, which is information indicating the event forecast announced during the data reference period, which is the reference period for the data used to forecast the event, and event performance information, which is information indicating the actual events that occurred during the data reference period, and generates an event forecasting error matrix, which is a matrix that enumerates in time series the error between the event forecast and the actual event that occurred during the data reference period, for each forecast target representing an event based on the event forecasting error matrix, and a vector whose elements are the average values ​​of each of the forecast targets obtained based on the event forecasting error matrix. The process involves generating an average value vector, generating a standard deviation vector whose elements are the standard deviations of each of the forecast targets, based on the event forecast error matrix, generating a correlation matrix based on the event forecast error matrix, which shows the correlations between the forecast targets, obtaining eigenvalues ​​and eigenvectors from the correlation matrix through eigenvalue analysis, identifying the principal eigenvalues ​​and eigenvectors, generating multiple forecast error vectors using random numbers whose elements are the errors of each of the forecast targets based on the average value vector, the standard deviation vector, and the principal eigenvalues ​​and eigenvectors, and generating multiple forecast error vectors for the forecast period obtained from the event forecast performance information to generate multiple event forecast scenarios.

[0009] Further issues disclosed in this application, and methods for solving them, will be made clear in the section on embodiments for carrying out the invention and in the drawings. [Effects of the Invention]

[0010] According to the present invention, forecasting scenarios for weather and other events can be generated efficiently and quickly. [Brief explanation of the drawing]

[0011] [Figure 1] This diagram shows a schematic configuration of a weather forecasting support system. [Figure 2] It is a diagram showing the main functions of the scenario generation device. [Figure 3A] It is an example of forecast error calculation information. [Figure 3B] It is an example of forecast error calculation information. [Figure 3C] It is an example of forecast error calculation information. [Figure 3D] It is an example of a weather forecast error matrix. [Figure 3E] It is an example of an average value vector. [Figure 3F] It is a diagram for explaining a method of generating a standard deviation vector. [Figure 3G] It is an example of a standard deviation vector. [Figure 3H] It is an example of a correlation matrix. [Figure 3I] It is an example of eigenvalues and eigenvectors. [Figure 3J] It is a diagram for explaining a method of calculating a forecast error vector. [Figure 3K] It is a diagram for explaining a method of generating a weather forecast scenario using a forecast error vector. [Figure 4] It is a system flowchart for explaining scenario generation processing. [Figure 5] It is an example of the hardware configuration of an information processing device used for realizing a scenario generation device or a scenario utilization device.

Embodiments for Carrying Out the Invention

[0012] Hereinafter, embodiments of the present invention will be described with reference to the drawings. In the following description, the character "S" attached before the reference numerals means processing steps.

[0013] FIG. 1 shows a schematic configuration of a weather forecast support system 1 (event forecast support system) shown as an embodiment of the present invention. As shown in the figure, the weather forecast support system 1 includes a device that generates a weather forecast scenario (hereinafter referred to as a "weather forecast scenario") (hereinafter referred to as a "scenario generation device 100"), a user terminal 2, an information providing device 6, and a device that uses the weather forecast scenario (hereinafter referred to as a "scenario use device 7").

[0014] The scenario generation device 100, the user terminal 2, the information providing device 6, and the scenario use device 7 are communicably connected via a communication network 5. The communication network 5 is a communication infrastructure that realizes communication by a wired or wireless method, and for example, is the Internet, a LAN (Local Area Network), a WAN (Wide Area Network), a wireless LAN, a power line communication network (Power Line Communication Network), various public communication networks, a dedicated line, or the like.

[0015] The information providing device 6 is an information processing device (computer), and for example, is an information providing server operated by an organization that provides weather information such as the Japan Meteorological Agency. The information providing device 6 provides the scenario generation device 100 with performance information of weather forecasts, performance information of weather, performance information of events related to weather such as typhoons and rainy seasons (hereinafter referred to as "weather events"), and the like.

[0016] The user terminal 2 is an information processing device (computer) operated by a user who uses the weather forecast support system 1. The user terminal 2 communicates with the scenario generation device 100 and the scenario use device 7 via the communication network 5, and for example, performs various settings for the scenario generation device 100 and the scenario use device 7, provides various information, and provides information generated by the scenario generation device 100 and the scenario use device 7 to the user. Note that the functions of the user terminal 2 may be provided to the scenario generation device 100 and the scenario use device 7 as well.

[0017] The scenario generation device 100 is an information processing device (computer) that generates multiple weather scenarios and provides them to the scenario utilization device 7. Based on actual weather forecast information and actual weather information (actual weather) provided by the information provision device 6, the scenario generation device 100 generates multiple weather forecast scenarios for a predetermined period in a predetermined area.

[0018] The scenario utilization device 7 is an information processing device (computer) that utilizes multiple weather forecast scenarios provided by the scenario generation device 100. The scenario utilization device 7 uses the provided multiple weather forecast scenarios, for example, to handle weather forecasts probabilistically / statistically. By utilizing multiple weather forecast scenarios, it makes probabilistic forecasts, such as "the probability that the temperature at the forecast date and time will be XX degrees or higher is XX% or higher." In this way, the scenario utilization device 7 can support various tasks that require probabilistic forecasting, such as probabilistic forecasting of electricity demand or probabilistic forecasting of electricity generation from renewable energy sources, performed by power companies and the like.

[0019] Figure 2 shows the main functions of the scenario generation device 100. As shown in the figure, the scenario generation device 100 has the following functions: a storage unit 110, a period setting reception unit 120, an information acquisition management unit 125, a forecast error calculation information generation unit 130, a weather forecast error matrix generation unit 135, an invalid period adjustment unit 137, an average value vector generation unit 140, a standard deviation vector generation unit 145, a correlation matrix generation unit 150, an eigenvalue analysis unit 155, a forecast error vector generation unit 160, and a weather forecast scenario generation unit 170.

[0020] Of the above functions, the memory unit 110 stores information (data) for the period setting information 101, weather forecast performance information 102 (event forecast performance information), weather performance information 103 (event performance information), weather event performance information 104, forecast error calculation information 105, weather forecast error matrix 106 (event forecast error matrix), mean vector 107, standard deviation vector 108, correlation matrix 109, eigenvalue / eigenvector 111, forecast error generation model 112, forecast error vector 113, and weather forecast scenario 114.

[0021] The period setting reception unit 120 receives from the user the period to be covered by the weather forecast (hereinafter referred to as the "forecast period") and the data reference period used to generate the forecast error calculation information 105 described later (hereinafter referred to as the "data reference period"). The period setting reception unit 120 receives the specification of the forecast period (e.g., 7 days, 14 days, etc.) and the data reference period (e.g., the most recent past 6 weeks (42 days) from the forecast creation date, etc.) via the user terminal 2, and manages the received forecast period and data reference period as period setting information 101 in the storage unit 110.

[0022] The Information Acquisition Management Unit 125 acquires weather forecast performance information 102, which includes weather forecasts announced by the Japan Meteorological Agency, etc., weather performance information 103, which includes actual weather information (weather actually observed), and weather event performance information 104, which includes information indicating the performance of weather events, and manages each acquired piece of information in the storage unit 110. The Information Acquisition Management Unit 125 acquires all or part of the weather forecast performance information 102, weather performance information 103, and weather event performance information 104 from, for example, the information providing device 6. The Information Acquisition Management Unit 125 also acquires all or part of this information from, for example, the user via the user terminal 2.

[0023] The forecast error calculation information generation unit 130 generates forecast error calculation information 105, which is information used to calculate the forecast error vector 113, based on the weather forecast performance information 102, weather performance information 103, and weather event performance information 104 (hereinafter also referred to as "basic information").

[0024] Figures 3A and 3C are examples of forecast error calculation information 105. Figure 3A shows forecast error calculation information 105 based on basic information in the case of no weather events (normal conditions). Figure 3B shows forecast error calculation information 105 based on basic information in the case of a weather event (approaching typhoon). Figure 3C shows forecast error calculation information 105 based on basic information in the case of a weather event (approaching typhoon, rainy season).

[0025] The forecast error calculation information 105 shown in Figures 3A to 3C consists of multiple records, each containing items such as elapsed time from 4 days after the forecast date (1051), forecast announcement date and time (1052), forecast target date and time (1053), predicted temperature (1054), predicted global solar radiation (1055), actual temperature (1056), actual global solar radiation (1057), typhoon approach flag (1058), and rainy season flag (1059). Each record of the forecast error calculation information 105 corresponds to a single forecast.

[0026] Of the above items, the elapsed time from 4 days after the forecast date (1051) stores the elapsed time from 4 days after the forecast announcement date and time. In this example, since a forecast with a 4-day time difference between the forecast announcement date and the forecast target date is used, the elapsed time is based on 4 days after the forecast announcement date and time. The forecast announcement date and time (1052) stores the date and time of the forecast announcement. The forecast target date and time (1053) stores the forecast target date and time. The predicted temperature (1054) stores the predicted temperature for the forecast target date and time. The predicted global solar radiation (1055) stores the predicted global solar radiation for the forecast target date and time. The actual temperature (1056) stores the temperature actually observed at the forecast target date and time. The actual global solar radiation (1057) stores the global solar radiation actually observed at the forecast target date and time. The typhoon approach flag 1058 stores a flag indicating whether or not a typhoon was approaching on the forecast date ("1" if approaching, "0" if not approaching). The rainy season flag 1059 stores a flag indicating whether or not the rainy season had started on the forecast date ("1" if the rainy season had started, "0" if not).

[0027] Returning to Figure 2, the weather forecast error matrix generation unit 135 calculates the weather forecast error (the difference between the weather forecast and the actually observed weather (temperature difference, humidity difference, etc.)) based on the forecast error calculation information 105, and manages the calculated error as the weather forecast error matrix 106.

[0028] Figure 3D shows an example of a weather forecast error matrix 106. As shown in the figure, the weather forecast error matrix 106 contains a list of weather forecast targets (hereinafter referred to as "forecast targets") in the column direction and the errors of the forecast targets on the day the weather forecast is issued by the Japan Meteorological Agency, etc. (the difference between the weather forecast and the actually observed weather, hereinafter referred to as "forecast error") in the row direction. The above forecast targets are, for example, target areas (Hiroshima, Okayama, Matsue, Tottori, Yamaguchi, etc.), target weather (temperature, humidity, solar radiation, etc.), and forecast times (every hour, every hour, etc.). In the example in the figure, there are only 5 areas, 3 weather types, and 24 × 7 time combinations (= 5 × 3 × 24 × 7 = 2520). In the following explanation, a vector whose elements are the values ​​(forecast errors) in each column of the weather forecast error matrix 106 will be referred to as a "forecast error vector".

[0029] Returning to Figure 2, the invalid period adjustment unit 137, in the eigenvalue analysis described later, excludes (invalidates) information from the weather forecast error matrix 106 generated by the weather forecast error matrix generation unit 135 during periods when the influence of solar radiation is small (solar radiation is below a predetermined threshold) (e.g., nighttime hours; hereinafter also referred to as the "invalid period"). The reason for excluding (invalidating) periods with a small influence of solar radiation from the weather forecast error matrix 106 in this way is that the error characteristics differ significantly when there is no solar radiation compared to when there is solar radiation. In the processing described later, which is performed after the eigenvalue analysis described later to restore the dimensionality of the forecast error vector, the invalid period adjustment unit 137 stores a predetermined value (for example, "0") in the element of the excluded period in the forecast error vector.

[0030] The average value vector generation unit 140 calculates the average value of the forecast error (the average value in the row direction of the weather forecast error matrix 106) for each forecast target in the weather forecast error matrix 106, generates a vector (hereinafter referred to as the "average value vector") whose elements are the calculated average values ​​of each forecast error, and manages the generated average value vector as the average value vector 107. Since the forecast error is affected by the weather observation method and time (season), it is preferable to set the data reference period to be as close as possible to the start date of the forecast period (for example, the most recent period). Figure 3E shows an example of the average value vector 107 based on the weather forecast error matrix 106.

[0031] Returning to Figure 2, the standard deviation vector generation unit 145 calculates the standard deviation for each forecast error in the weather forecast error matrix 106, generates a vector (hereinafter referred to as the "standard deviation vector") whose elements are the calculated standard deviations of each forecast error, and manages the generated standard deviation vector as the standard deviation vector 108. Here, if there is a weather event such as an approaching typhoon or the start of the rainy season, the forecast error will behave differently than usual, so the standard deviation vector generation unit 145 considers the weather event (the flags related to the weather event in the weather forecast error matrix 106 (typhoon approach flag, rainy season flag, etc.; hereinafter referred to as the "weather event flag")) and generates different standard deviation vectors for normal times and when a weather event is present.

[0032] Figure 3F illustrates the method for generating the standard deviation vector 108. The upper part of the figure shows an example of forecast error calculation information 105, and the lower part of the figure shows the standard deviation values ​​(elements of the standard deviation vector) for each weather event based on the forecast error calculation information 105 in the upper part of the figure. The values ​​for each weather event (normal, approaching typhoon, start of rainy season, end of rainy season, etc.) in the table in the lower part of the figure become elements of the standard deviation vector 108 for each event. Figure 3G shows an example of the standard deviation vector 108 when there are no weather events (normal conditions).

[0033] Returning to Figure 2, the correlation matrix generation unit 150 normalizes the weather forecast error matrix 106 and generates a correlation matrix 109 between forecast targets based on the normalized weather forecast error matrix 106. Here, the normalization of the weather forecast error matrix 106 is performed, for example, using the mean vector and standard deviation vector of the forecast error obtained using data without weather events (normal conditions) within the data reference period (i.e., a correlation matrix of forecast targets during normal conditions without weather events is generated). This makes it possible to exclude the influence of large forecast errors due to weather events. Note that the standard deviation vector is set separately, so there is no need to specifically grasp the correlation matrix that includes information on weather events.

[0034] Figure 3H shows an example of a correlation matrix 109. As shown in the figure, the correlation matrix lists the forecast targets in the same order in the column and row directions, and includes the correlation coefficient (0 to 1) at the corresponding (intersection) position between the forecast target in each row and the forecast target in each column.

[0035] Returning to Figure 2, the eigenvalue analysis unit 155 obtains eigenvalues ​​and eigenvectors by performing eigenvalue analysis on the correlation matrix 109 excluding the invalid period, and manages the obtained eigenvalues ​​and eigenvectors as eigenvalue / eigenvector 111 (hereinafter, eigenvalues ​​and eigenvectors together will also be referred to as "features"). The eigenvalue analysis unit 155 also performs dimensionality reduction of the features of the correlation matrix 109 by evaluating the explanatory contribution rate, which will be described below.

[0036] Figure 3I shows an example of principal components selected by the eigenvalue analysis unit 155. The figure shows the eigenvalues ​​("Number 1" to "Number 8") and eigenvectors of the principal components, as well as the explanatory contribution rate of each eigenvector. The user can set a threshold for the explanatory contribution rate used for selecting principal components (selection of features) (the lower limit of the explanatory contribution rate to be adopted as a principal component) in the scenario generation device 100 via the user terminal 2.

[0037] Returning to Figure 2, the forecast error vector generation unit 160 inputs the mean vector 107, standard deviation vector 108, and eigenvalue / eigenvector 111 (principal components) generated as described above into the forecast error generation model 112, which is represented by the following equation. This generates a number of independent forecast error vectors using random numbers ε, and manages the generated forecast error vectors as forecast error vectors 113.

[0038]

number

[0039] In the above equation, the mean vector in the first term on the right-hand side is the mean vector 107 generated by the mean vector generation unit 140. The standard deviation vector in the second term on the right-hand side is the standard deviation vector 108 generated by the standard deviation vector generation unit 145. As mentioned above, the standard deviation vector 108 is selected to correspond to the weather events during the period covered by the weather forecast. The value of M in the summation symbol (Σ) of the second term on the right-hand side is the number of features determined based on the threshold (lower limit) of the explanatory contribution rate, and is an eigenvalue. i and eigenvectors i These are the eigenvalues ​​and eigenvectors obtained by the eigenvalue analysis unit 155. ε i This is an automatically generated random number (for example, a random number generated according to a normal distribution N(0,1)).

[0040] Figure 3J is a diagram illustrating in more detail the method for calculating the forecast error vector (Equation 1). In this figure, N is the number of targets to be forecasted. As shown in this figure, the forecast error of the nth element of the forecast error vector can be obtained from the following equation.

[0041]

number

[0042] Returning to Figure 2, the weather forecast scenario generation unit 170 generates multiple weather forecast scenarios by applying multiple forecast error vectors 113 (with solar radiation during the invalid period added) generated by the forecast error vector generation unit 160 to the weather forecast for the forecast period obtained from the weather forecast performance information 102, and stores the generated multiple weather forecast scenarios as weather forecast scenario 114 in the storage unit 110.

[0043] Figure 3K shows an example of generating a weather forecast scenario 114 by applying a forecast error vector 113 to the weather forecast for the forecast period. As shown in the figure, in this example, multiple weather forecast scenarios are generated by adding the forecast error to the weather (temperature, solar radiation) of the weather forecast for the forecast period (for example, a weather forecast obtained from the Japan Meteorological Agency) obtained from the weather forecast performance information 102.

[0044] Next, we will explain the main processes performed by the scenario generation device 100, which has the above configuration (hereinafter referred to as "scenario generation process S400").

[0045] Figure 4 is a system flow diagram illustrating the scenario generation process S400. The scenario generation process S400 will be explained below in conjunction with this diagram.

[0046] First, the period setting reception unit 120 receives the setting of the forecast period and data reference period from the user via the user terminal 2, and stores the received information as period setting information 101 in the storage unit 110 (S411).

[0047] Next, the information acquisition and management unit 125 acquires weather forecast performance information 102, weather performance information 103, and weather event performance information 104 corresponding to the forecast period and data reference period, respectively, and stores each of the acquired pieces of information in the storage unit 110.

[0048] Next, the forecast error calculation information generation unit 130 generates forecast error calculation information 105 based on the weather forecast performance information 102, weather performance information 103, and weather event performance information 104 (basic information) (S412). As mentioned above, the forecast error calculation information generation unit 130 generates forecast error calculation information 105 separately for cases where there are no weather events (normal) and for cases where there are weather events (weather event present).

[0049] Next, the weather forecast error matrix generation unit 135 generates a weather forecast error matrix based on the forecast error calculation information 105 (S413), and the invalid period adjustment unit 137 generates a weather forecast error matrix 106 by excluding the solar radiation error for a specified period (excluding the invalid period) from the weather forecast error matrix generated by the weather forecast error matrix generation unit 135 (S416). The invalid period adjustment unit 137 also receives, for example, a user via the user terminal 2 to specify conditions to be used in calculating the invalid period (for example, "the period during which the amount of solar radiation is below a predetermined threshold will be considered an invalid period," etc. Hereinafter referred to as "invalid conditions") (S414), and calculates the invalid period based on the received invalid conditions (S415).

[0050] Next, the mean value vector generation unit 140 generates a mean value vector and a standard deviation vector based on the weather forecast error matrix 106, and stores the generated mean value vector as mean value vector 107 and the generated standard deviation vector as standard deviation vector 108 in the storage unit 110 (S417).

[0051] Furthermore, the correlation matrix generation unit 150 generates a correlation matrix based on the weather forecast error matrix 106 excluding the invalid period, and stores it in the storage unit 110 as the correlation matrix 109 (S421).

[0052] Next, the eigenvalue analysis unit 155 generates eigenvalues ​​and eigenvectors by performing eigenvalue analysis on the correlation matrix 109, stores the generated eigenvalues ​​and eigenvectors as eigenvalue / eigenvector 111 in the memory unit 110, and further performs explanatory contribution evaluation to obtain the eigenvalues ​​and eigenvectors of the principal components. In other words, the eigenvalue analysis unit 155 reduces the number of features (S422). The correlation matrix 109 that is the target of eigenvalue analysis is the weather forecast error matrix 106 (under normal conditions) with invalid periods excluded.

[0053] Next, the forecast error vector generation unit 160 generates multiple forecast error vectors by applying the mean value vector 107, the standard deviation vector 108, and the eigenvalue / eigenvector 111 (principal components) to the forecast error generation model 112 (S431). At this point, the forecast error vector is based on the normal weather forecast error matrix 106.

[0054] Next, the forecast error vector generation unit 160 adds an element to each of the multiple forecast error vectors generated in S431 that sets the forecast error for solar radiation during the invalid period to "0", thereby generating multiple forecast error vectors with the same number of elements as the forecast period before the invalid period was excluded, and stores the generated multiple forecast error vectors as forecast error vector 113 in the storage unit 110 (S432).

[0055] Next, the weather forecast scenario generation unit 170 generates multiple weather forecast scenarios (S433) by applying multiple forecast error vectors 113 generated by the forecast error vector generation unit 160 to the weather forecast for the forecast period obtained from the weather forecast performance information 102, and stores the generated multiple weather forecast scenarios as weather forecast scenario 114 in the storage unit 110. The scenario generation device 100 provides the weather forecast scenario 114 to the scenario utilization device 7 as needed, for example, by transmitting it via the communication network 5.

[0056] As described above, the weather forecast support system 1 of this embodiment calculates the mean and standard deviation of the error between the weather forecast and the actually observed weather (forecast error) for each forecast target, calculates eigenvalues ​​and eigenvectors based on the correlation matrix of the forecast target, and further evaluates the explanatory contribution rate to reduce the dimensionality of the features that affect the correlation of the forecast target.Then, the weather forecast support system 1 uses the mean and standard deviation for each forecast target, and generates multiple pseudo-errors (forecast error vectors) using random numbers while reflecting the dimensionality-reduced features (eigenvalues ​​and eigenvectors), and generates multiple weather forecast scenarios by reflecting the generated errors in the weather forecast for the forecast period.In this way, the weather forecast support system 1 of this embodiment generates forecast error vectors using principal components identified by eigenvalue analysis and generates multiple weather forecast scenarios using the generated forecast error vectors, so it can generate multiple weather forecast scenarios efficiently and quickly with a small computational load.

[0057] Furthermore, the weather forecasting support system 1 generates standard deviation vectors individually according to weather events, and uses the standard deviation vectors corresponding to weather events during the forecast period as the standard deviation vectors used to generate the forecast error vector. Therefore, it can generate weather forecasting scenarios that take into account the impact of weather events.

[0058] Furthermore, the weather forecasting support system 1 excludes (invalidates) information from the weather forecasting error matrix for periods where the influence of solar radiation is small, as the error characteristics differ significantly compared to when solar radiation is present. This allows for efficient eigenvalue analysis.

[0059] <Example of an information processing device> Figure 5 shows an example of the hardware configuration of an information processing device used to realize the scenario generation device 100 and the scenario utilization device 7. The illustrated information processing device 10 comprises a processor 11, main memory 12, auxiliary memory 13, input device 14, output device 15, and communication device 16. Specific examples of the information processing device 10 include, for example, personal computers, office computers, various server devices, and general-purpose computers. The information processing device 10 may be implemented, in whole or in part, using virtual information processing resources provided using virtualization technology, such as a virtual server provided by a cloud system. The scenario generation device 100 and the scenario utilization device 7 may be implemented using multiple information processing devices 10 that are connected to each other in a communicative manner.

[0060] In the figure, the processor 11 is composed of, for example, a CPU (Central Processing Unit), an MPU (Micro Processing Unit), a GPU (Graphics Processing Unit), an FPGA (Field Programmable Gate Array), an ASIC (Application Specific Integrated Circuit), an AI (Artificial Intelligence) chip, etc.

[0061] The main memory 12 is a device for storing programs and data, and is, for example, ROM (Read Only Memory), RAM (Random Access Memory), or non-volatile memory (NVRAM (Non-Volatile RAM)).

[0062] The auxiliary storage device 13 includes, for example, an SSD (Solid State Drive), a hard disk drive, an optical storage device (CD (Compact Disc), DVD (Digital Versatile Disc), etc.), a storage system, an IC card, a reader / writer for recording media such as SD cards and optical recording media, and the storage area of ​​a cloud server. Programs and data can be read into the auxiliary storage device 13 via a recording media reader or a communication device 16. Programs and data stored in the auxiliary storage device 13 are read into the main memory 12 as needed.

[0063] The input device 14 is an interface that accepts input from an external source, and can be, for example, a keyboard, mouse, touch panel, card reader, pen-input tablet, or voice input device.

[0064] The output device 15 is an interface that outputs various information such as processing progress and processing results. The output device 15 may be, for example, a display device that visualizes the above information (LCD (Liquid Crystal Display), graphics card, etc.), a device that converts the above information into sound (speaker, etc.), or a device that converts the above information into text (printer, etc.). For example, the information processing device 10 may be configured to input and output information to and from other devices via the communication device 16.

[0065] The input device 14 and the output device 15 constitute a user interface for receiving and presenting information with the user.

[0066] The communication device 16 is a device that enables communication (wired or wireless communication) with other devices via a communication infrastructure such as the communication network 5, and is configured using, for example, a NIC (Network Interface Card), a wireless communication module, a USB module, etc.

[0067] The information processing device 10 may have, for example, an operating system, a file system, a DBMS (Database Management System) (relational database, NoSQL, etc.), a KVS (Key-Value Store), etc. installed on it.

[0068] The functions of the scenario generation device 100 and the scenario utilization device 7 are realized either by the processor 11 of the information processing device 10 reading and executing a program stored in the main memory 12, or by the functions of the hardware (FPGA, ASIC, AI chip, etc.) that constitutes the scenario generation device 100 and the scenario utilization device 7 themselves. The scenario generation device 100 and the scenario utilization device 7 store the aforementioned various types of information (data) as, for example, database tables or files managed by a file system.

[0069] The embodiments of the present invention have been described in detail above, but this description is for the purpose of facilitating understanding of the present invention and does not limit it. The present invention can be modified and improved without departing from its spirit, and of course, equivalents thereof are included in the present invention. For example, the above embodiments have been described in detail for the purpose of explaining the present invention in an easy-to-understand manner and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to add, delete, or replace some of the configurations of the above embodiments with other configurations.

[0070] For example, the present invention not only uses the eigenvalues ​​and eigenvectors of the principal components obtained by eigenvalue analysis, but also allows for the individual setting of mean vectors and standard deviation vectors. This configuration of the present invention can be broadly applied to predicting fluctuations in events other than weather (such as stock prices, exchange rates, interest rates, commodity prices, GDP (Gross Domestic Product), etc.). [Explanation of Symbols]

[0071] 1. Weather forecasting support system 2 User terminals 5. Communication Network 6 Information provision device 7 Scenario-based device 100 Scenario Generator 110 Storage section 101 Period Setting Information 102 Weather forecast performance information 103 Weather Information 104 Weather Event History Information 105 Forecast Error Calculation Information 106 Weather forecast error matrix 107 Mean Vector 108 Standard Deviation Vectors 109 Correlation Matrix 111 Eigenvalues / Eigenvectors 112 Forecast Error Generation Model 113 Forecast Error Vector 114 Weather Forecast Scenarios 120 Period Setting Reception Department 125 Information Acquisition Management Department 130 Forecast Error Calculation Information Generation Unit 135 Weather forecast error matrix generation unit 137 Invalid Period Adjustment Unit 140 Mean Value Vector Generation Unit 145 Standard Deviation Vector Generation Unit 150 Correlation Matrix Generation Unit 155 Eigenvalue Analysis Department 160 Forecast Error Vector Generation Unit 170 Weather Forecast Scenario Generation Unit S400 Scenario Generation Process

Claims

1. It is configured using an information processing device having a processor and memory, The forecast period is the period covered by the forecast for the event, Event forecast performance information includes event forecasts, which are information indicating event forecasts announced during the data reference period, which is the reference period for data used to forecast events, and Event performance information, which includes information showing the actual events that occurred during the aforementioned data reference period, Remember this, Based on the aforementioned event forecast performance information and the aforementioned event performance information, an event forecast error matrix is ​​generated, which is a matrix that enumerates in time series the errors between the event forecast and the event that actually occurred during the data reference period for each forecast target representing an event. Based on the aforementioned event forecast error matrix, an average value vector is generated, which is a vector whose elements are the average values ​​of each of the forecast targets. Based on the aforementioned event forecast error matrix, a standard deviation vector is generated, which is a vector whose elements are the standard deviations of each of the forecast targets. Based on the aforementioned event forecast error matrix, a correlation matrix is ​​generated, which is a matrix showing the correlation between the forecast targets. Eigenvalues ​​and eigenvectors are obtained from the aforementioned correlation matrix by eigenvalue analysis. Identify the eigenvalues ​​and eigenvectors of the principal components, Based on the mean vector, the standard deviation vector, and the eigenvalues ​​and eigenvectors of the principal components, multiple forecast error vectors are generated using random numbers, each of which has the error of the forecast target as an element. Multiple event forecast scenarios are generated by reflecting the generated multiple forecast error vectors into the event forecast for the forecast period obtained from the event forecast performance information. Event forecasting support system.

2. An event prediction support system according to claim 1, The aforementioned event is meteorological. Event forecasting support system.

3. The event prediction support system according to claim 2, Further, the system stores weather event history information, which indicates whether or not a weather event occurred. Based on the aforementioned weather event performance information, the standard deviation vector corresponding to the weather event is generated, The standard deviation vector used to generate the forecast error vector is the standard deviation vector corresponding to the weather event during the forecast period. Event forecasting support system.

4. The event prediction support system according to claim 2, The aforementioned event performance information includes information indicating solar radiation, When determining the eigenvalues ​​and eigenvectors of the correlation matrix, elements of the event prediction error matrix whose solar radiation is less than or equal to a predetermined threshold are excluded. Event forecasting support system.

5. The event prediction support system according to claim 2, The forecast subject is a combination of the area covered by the weather forecast, the weather subject to the weather forecast, and the date and time within the forecast period. Event forecasting support system.

6. The event prediction support system according to claim 2, The aforementioned eigenvalue analysis includes a user interface that accepts the setting of a lower limit for the explanatory contribution rate used when selecting principal components. Event forecasting support system.

7. An information processing device having a processor and memory, The forecast period is the period covered by the forecast for the event, Event forecast performance information includes event forecasts, which are information indicating event forecasts announced during the data reference period, which is the reference period for data used to forecast events, and Event performance information, which includes information showing the actual events that occurred during the aforementioned data reference period, Steps to memorize, Based on the aforementioned event forecast performance information, a step is to generate an event forecast error matrix, which is a matrix that enumerates in time series the errors between the event forecast and the event that actually occurred during the data reference period for each forecast target representing an event. A step of generating an average value vector, which is a vector whose elements are the average values ​​of each of the forecast targets, obtained based on the event forecast error matrix. A step of generating a standard deviation vector, which is a vector whose elements are the standard deviations of each of the forecast targets, obtained based on the event forecast error matrix; A step of generating a correlation matrix, which is a matrix showing the correlation between the forecast targets, based on the event forecast error matrix. The steps include: obtaining eigenvalues ​​and eigenvectors from the aforementioned correlation matrix using eigenvalue analysis; Identify the eigenvalues ​​and eigenvectors of the principal components, The steps include generating multiple forecast error vectors using random numbers, each of which is a vector whose elements are the errors of the forecast target, based on the mean vector, the standard deviation vector, and the eigenvalues ​​and eigenvectors of the principal components, and A step of generating multiple event forecast scenarios by reflecting the generated multiple forecast error vectors in the event forecast for the forecast period obtained from the event forecast performance information, A method for supporting event forecasting, which performs the following actions.

8. The event prediction support method according to claim 7, The aforementioned event is meteorological. Event forecasting support method.

9. The event prediction support method according to claim 8, The aforementioned information processing device A further step involves storing weather event history information, which indicates whether or not a weather event occurred. A step of generating the standard deviation vector corresponding to the weather event based on the weather event performance information, The step of using the standard deviation vector corresponding to the weather event during the forecast period as the standard deviation vector used to generate the forecast error vector, A method for supporting event forecasting, which further implements the above.

10. The event prediction support method according to claim 8, The aforementioned event performance information includes information indicating solar radiation, The aforementioned information processing device When determining the eigenvalues ​​and eigenvectors of the correlation matrix, the step of excluding elements of the event prediction error matrix where the solar radiation is less than or equal to a predetermined threshold, A method for supporting event forecasting, which further implements the above.

11. The event prediction support method according to claim 8, The forecast subject is a combination of the area covered by the weather forecast, the weather subject to the weather forecast, and the date and time within the forecast period. Event forecasting support method.

12. The event prediction support method according to claim 8, The system includes a user interface that accepts the setting of a lower limit for the explanatory contribution rate used when selecting principal components in the aforementioned eigenvalue analysis. Event forecasting support method.

Images