Information processing device, information processing method, and information processing program

Abstract

To provide an information processing device, an information processing method, and an information processing program which suppress deterioration in estimation accuracy in an estimation value of an estimation object.SOLUTION: In an information processing system 1, a processing section 18 of an information processing device 10 includes: an estimation section 18D for calculating a first estimation value at a first time of an estimation object according to a plurality of third actual values at a third time in a second predetermined period which includes a period before a second time in a first predetermined period back from the first time; and a correction section 18E for correcting the first estimation value according to the degree of deviation between a second estimation value at the second time and a second actual value.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] An embodiment of the present invention relates to an information processing device, an information processing method, and an information processing program. [Background technology]

[0002] Systems are known that estimate values ​​that vary with date and time. For example, systems are known that estimate a trading price, a bid amount, etc. at a certain point in time in an electricity trading market. Specifically, since the demand and supply of electricity are mainly influenced by meteorological factors, a system has been disclosed that utilizes meteorological estimation by a numerical weather simulator to estimate the electricity market price, the bid amount, etc. on a target date by using actual values ​​of the electricity market price on a number of past dates that are similar to the meteorological conditions on the target date. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-149475 A [Patent Document 2] International Publication No. 2016 / 136323 Summary of the Invention [Problem to be solved by the invention]

[0004] However, in conventional technology, estimation is performed using the average of actual values ​​from multiple past days with similar weather conditions, so it is not possible to respond to fluctuations in actual values ​​closer to the target estimation date, such as immediately before the target estimation date, and estimation accuracy may be reduced.

[0005] An object of the present disclosure is to provide an information processing device, an information processing method, and an information processing program capable of suppressing a decrease in estimation accuracy of an estimation value of an estimation target. [Means for solving the problem]

[0006] According to an embodiment, an information processing device includes a processor that calculates a first estimate value at a first time based on third actual values ​​at a plurality of third times within a second predetermined period that includes a period going back from a first time to be estimated and prior to a second time within a first predetermined period, and corrects the first estimate value based on a degree of deviation between the second estimate value at the second time and the second actual value. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a schematic diagram of an information processing system according to an embodiment. [Diagram 2] 5 is an explanatory diagram of information relating to time handled by the information processing apparatus of the embodiment. [Diagram 3] Schematic diagram of the data structure of the weather estimation DB. [Figure 4] Schematic diagram of the data structure of the performance DB. [Diagram 5] FIG. [Figure 6] 1 is a flowchart showing the flow of information processing. [Figure 7] Schematic diagram of data used in the operation example. [Figure 8] Schematic diagram of data used in the operation example. [Figure 9] Schematic diagram of calculation results of an operation example. [Figure 10] Schematic diagram of calculation results of an operation example. [Figure 11] Schematic diagram of data used in the operation example. [Figure 12] Schematic diagram of calculation results of an operation example. [Figure 13] Schematic diagram of calculation results of an operation example. [Figure 14] Schematic diagram of calculation results of an operation example. [Figure 15] Schematic diagram of calculation results of an operation example. [Figure 16] Hardware configuration diagram. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] An information processing apparatus, an information processing method, and an information processing program according to embodiments will be described in detail below with reference to the accompanying drawings.

[0009] In the following, the time can refer to two types of time: a moment in time and a time span of a unit time. In this embodiment, the time is used to represent the time span of a unit time as an example. The unit time is, for example, 30 minutes, but is not limited to this unit.

[0010] FIG. 1 is a schematic diagram showing an example of an information processing system 1 according to the present embodiment.

[0011] The information processing system 1 is a system that estimates an estimated value of an estimation target that varies depending on date and time. The estimated value is not an actual value but an estimated value. In this embodiment, a form in which the information processing system 1 estimates an electricity market price at a certain time in the electricity trading market is described as an example. In this embodiment, a time may be described as a time frame to indicate that the time represents a time span. The electricity market price is a trading price per unit time of buying and selling in the electricity trading market. That is, in this embodiment, a case in which the estimated value is an estimated value of the electricity market price per unit time of buying and selling in the electricity trading market is described as an example. Electricity market prices include a spot market price, an hourly market price, and an imbalance charge price.

[0012] The estimated value and the actual value described later may be values ​​that may vary over time such as date and time, and are not limited to the electricity market price per unit time in the electricity trading market. For example, the estimated value and the actual value may be, for example, the trading price of electricity traded at the electricity exchange, the bid amount, the electricity demand amount, the power generation amount of a renewable energy source, etc. The estimated value and the actual value may be the sales amount, the sales number, the shipment amount, the production amount, the amount of congestion of people and vehicles, the congestion degree, the flow rate of people, etc. of a product. In this embodiment, a first estimated value, a second estimated value, and a third estimated value are used as the estimated value. The first estimated value is an estimated value at the first time T1. The second estimated value is an estimated value at the second time T2. The third estimated value is an estimated value at the third time T3. In this embodiment, the estimated value and the actual value are described as being the electricity market price per unit time of buying and selling electricity. Details of the first time T1, the second time T2, and the third time T3 will be described later.

[0013] The information processing system 1 includes a weather estimation system 2, a performance value management system 3, and an information processing device 10. The weather estimation system 2, the performance value management system 3, and the information processing device 10 are communicatively connected via a network NW or the like.

[0014] The weather estimation system 2 is a system that estimates meteorological information (meteorological quantities) at a plurality of points in the power supply area of ​​a power company. In the case of Japan, the power supply areas include Hokkaido, Tohoku, Tokyo, Chubu, Hokuriku, Kansai, Shikoku, Chugoku, Kyushu, etc. The points are, for example, installation points of observation stations (AMeDAS: Automated Meterological Data Acquisition System) of the Japan Meteorological Agency. The weather estimation system 2 is an external system of the information processing device 10. The weather estimation system 2 is, for example, a meteorological information server, web service, cloud service, etc. provided by domestic or foreign government agencies such as the Japan Meteorological Agency, private businesses, or international organizations.

[0015] The weather estimation system 2 estimates weather information based on the observed values ​​of weather information from observation stations at each point. For example, the weather estimation system 2 estimates weather information using a Weather Research and Forecasting model (WRF) or the like. The weather information includes, for example, temperature, solar radiation, sunshine hours, humidity, wind speed, wind direction, precipitation, etc. Solar radiation may be called solar radiation intensity. The weather estimation system 2 associates a weather estimation value, which is the value of the estimated weather information, with an ID (Identification) of a location and time information (date and time), and stores the associated weather estimation data in an internal database (DB). The weather estimation value has a numerical format, for example. In response to a request from the information processing device 10, the weather estimation system 2 provides the weather estimation data stored in the DB to the information processing device 10.

[0016] The performance value management system 3 holds performance data of the estimation target. The performance value management system 3 is an external system of the information processing device 10. The performance data is data in which a performance value and a time at which the performance value was obtained are associated with each other. The performance value is an actual value, that is, a past value. As with the above-mentioned estimated value, examples of the performance value include the electricity market price of electricity traded at the electricity exchange, the bid amount, the electricity demand amount, the power generation amount of a renewable energy power source, the sales amount of a product, the number of sales, the shipment amount, the production amount, the amount of congestion of people and vehicles, the degree of congestion, the flow rate of people, and the like. In this embodiment, a first performance value, a second performance value, and a third performance value are used as the performance value. The first performance value is the performance value at the first time T1. The second performance value is the performance value at the second time T2. The third performance value is the performance value at the third time T3. In this embodiment, as described above, the performance value is described as the electricity market price per unit time of buying and selling electricity.

[0017] Here, electricity market prices will be explained using the Japan Electric Power Exchange (JEPX) as an example. In the spot (one day ahead) market set up on JEPX, a day is divided into 48 time slots based on the measurement unit of electricity (i.e., from 0 minutes to 30 minutes and from 30 minutes to 60 minutes of each hour), and electricity is traded by time slot. The unit of trading power volume is 1MW (30 minutes), which is 500kWh in terms of power volume. Bid prices can be specified in increments of 0.01 yen per kWh. In the spot market, bidding for the next day's time slots closes at 10am every day.

[0018] Each time slot in the spot market is independent of the others, and the contract price and contract quantity are determined for each time slot using a blind single-price auction system.

[0019] Depending on the results of spot market contracts, there may be cases where the limit on the amount of electricity that can be passed through the interconnection lines of the power grid is exceeded. For example, the intersection point between the buy (demand) curve, which is a combination of bids from all over Japan, and the sell (supply) curve, which is a combination of bids from all over Japan, is found, and trading is established at the price and quantity at that intersection point. In this case, depending on the interconnection lines between areas, the amount of electricity that can be transmitted may be small, so the upper limit of the interconnection line may be exceeded. In such a time frame where the upper limit is exceeded, the area is divided at the interconnection line. Then, the intersection point of the demand curve and the supply curve is found independently in each divided area, and the contract price and contract quantity are determined. This state is called market division. At this time, the price agreed upon throughout Japan is called the system price, and the price agreed upon in each area is called the area price.

[0020] The power market price managed by the performance value management system 3 may be either a system price or an area price. In this embodiment, the performance value management system 3 manages performance data of the power market price for each area where the power market price is determined. The performance value management system 3 may obtain performance data of the power market price from the power exchange system or another system linked to the power exchange system, or the performance value management system 3 may be a part of the power exchange system.

[0021] Next, a description will be given of the information processing device 10. The information processing device 10 uses weather estimation data acquired from the weather estimation system 2 and performance data acquired from the performance value management system 3 to execute information processing such as correction of a first estimated value at a first time T1 of an estimation target.

[0022] FIG. 2 is an explanatory diagram of information relating to time handled by the information processing device 10 of this embodiment.

[0023] The information processing device 10 of this embodiment handles times such as a first time T1, a second time T2, and a third time T3. These times represent a time span of a unit time, that is, a time frame. In this embodiment, a case where these times are a time span of 30 minutes, which is a time frame handled in the spot market, is described as an example. That is, these times are represented by year, month, day, and time frames.

[0024] The first time T1 is the time of the estimation target. In detail, the first time T1 is a specific time frame on the date of the estimation target of the estimated value. In other words, the first time T1 is the date and time frame of the estimation target. The first time T1 may be a time frame of any timing, including the present, past, and future. The past is a period in which an actual value is obtained. The present is a time when the latest actual value is obtained. The future is a time when an actual value is not obtained. In this embodiment, a case in which the first time T1 is a time frame of a future timing will be described as an example. The specific date and time frame of the first time T1 is specified by an operation instruction by the user, etc.

[0025] The second time T2 is a time within the first predetermined period Q1 going back from the first time T1. Going back from the first time T1 means including a time frame immediately preceding the first time T1 in a direction toward the past relative to the first time T1, and including one or more time frames consecutive from the time frame in a direction toward the past. "Consecutive" means "adjacent."

[0026] The first predetermined period Q1 may be any period including at least one time frame, and is a period that is specified and changed according to an operation instruction by a user or the like. The first predetermined period Q1 is a period during which it is highly likely that an actual measurement value substantially the same as that at the first time T1 is obtained. Specifically, the first predetermined period Q1 is a period immediately before the first time T1.

[0027] When the first predetermined period Q1 is a period of one time frame, the second time T2 means a time frame immediately before the first time T1, which is the date and time frame to be estimated. When the first predetermined period Q1 is a period of one day, the second time T2 means a time frame the same as the first time T1, which is included in a period of one day immediately before the first time T1, which is the date and time frame to be estimated. The same time frame means a time span of the same time. Specifically, when the first time T1 is a time frame from 10:00 to 10:30 a.m., for example, the same time frame means a time frame from 10:00 to 10:30 a.m. on a different day.

[0028] In this embodiment, a case will be described as an example in which the first predetermined period Q1 is seven days immediately before the first time T1. Note that the first predetermined period Q1 can be changed as appropriate by an operational instruction by the user or the like.

[0029] The third time T3 is a plurality of times within the second predetermined period Q2 including a period before the first time T1 and the second time T2. The period before the first time T1 and the second time T2 means a period including at least a time frame before each of the time frame one time before the first time T1 and the time frame one time before the second time T2. FIG. 2 shows an example in which the second predetermined period Q2 is a period before any second time T2 within the first predetermined period Q1, includes a period before the most recent second time T2 within the first predetermined period Q1, and includes a plurality of third times T3. The second predetermined period Q2 may be a second predetermined period Q2 period longer than the first predetermined period Q1 going back from a time frame one time before the first time T1.

[0030] The length of the second predetermined period Q2 is a period including a plurality of time frames. In addition, the length of the second predetermined period Q2 is preferably longer than the length of the first predetermined period Q1. The second predetermined period Q2 is, for example, 3 months, 6 months, 1 year, etc., but is not limited to these periods. In addition, the second predetermined period Q2 can be appropriately changed by an operational instruction by the user, etc.

[0031] Returning to Figure 1, we will continue with the explanation.

[0032] The information processing device 10 includes a communication unit 12, a UI (user interface) unit 14, a storage unit 16, and a processing unit 18. The communication unit 12, the UI unit 14, and the storage unit 16 are communicatively connected to the processing unit 18 via a bus or the like.

[0033] The communication unit 12 communicates with an external information processing device via a network NW or the like. In this embodiment, the communication unit 12 communicates at least with the weather estimation system 2 and the performance value management system 3. The UI unit 14 has an input function for accepting operation input by a user and an output function for outputting various information. The input function is, for example, an input device such as a keyboard. The output function is, for example, a display for displaying various information, a speaker for outputting sound, etc. The memory unit 16 stores various information.

[0034] The processing unit 18 is a calculation unit that executes information processing, and includes a processing information acquisition unit 18A, a second estimated value acquisition unit 18B, a deviation calculation unit 18C, an estimation unit 18D, a correction unit 18E, and an output control unit 18F.

[0035] At least one of the processing information acquisition unit 18A, the second estimated value acquisition unit 18B, the deviation calculation unit 18C, the estimation unit 18D, the correction unit 18E, and the output control unit 18F is realized, for example, by one or more processors. For example, each of the above units may be realized by having a processor such as a CPU execute a program, that is, by software. Each of the above units may be realized by a processor such as a dedicated IC (Integrated Circuit), that is, by hardware. Each of the above units may be realized by using both software and hardware. When multiple processors are used, each processor may realize one of the units, or may realize two or more of the units.

[0036] Furthermore, at least one of the above-mentioned functional units included in the processing unit 18 and at least a part of the various pieces of information stored in the storage unit 16 may be mounted on an external information processing device communicatively connected to the information processing device 10. Specifically, the flow rate estimation unit 20 included in the processing unit 18 may be provided in an external information processing device connected to the information processing device 10 via a network.

[0037] The processing unit 18 may be realized by, for example, one or more processing circuits such as a CPU, a microprocessor, a GPU (Graphics Processing Unit), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or an electronic circuit including these circuits, etc. The processing unit 18 may also be realized by an information processing device such as a computer, a computer system configured by multiple computers or servers communicating with each other via a network, or a PC cluster in which multiple computers cooperate to execute information processing.

[0038] The processing information acquisition unit 18A acquires processing information used for processing by the processing unit 18.

[0039] The processing information includes weather estimation data, performance data, hyperparameters, a first time T1 which is the date and time frame of the estimation target, and an estimation target area.

[0040] The processing information acquisition unit 18A acquires weather estimation data from the weather estimation system 2. The processing information acquisition unit 18A transmits an acquisition request for weather estimation data to the weather estimation system 2, acquires weather estimation data including a location ID, a weather estimation value, and time information (date and time) of each location from the weather estimation system 2, and stores the data in the storage unit 16. For example, the processing information acquisition unit 18A may transmit an acquisition request for weather estimation data to the weather estimation system 2 at regular intervals and acquire unacquired weather estimation data. Alternatively, the processing information acquisition unit 18A may specify an acquisition target period and acquire weather estimation data belonging to the specified period from the weather estimation system 2. The processing information acquisition unit 18A may also acquire weather estimation data by a push notification from the weather estimation system 2.

[0041] The processing information acquisition unit 18A registers the weather estimation data acquired from the weather estimation system 2 in a weather estimation DB (database) 16A of the storage unit 16.

[0042] Fig. 3 is a schematic diagram of an example of the data configuration of the weather estimation DB 16A. The weather estimation DB 16A is a database in which weather estimation data is registered in which the date and time are associated with the weather estimation value for each area name. As shown in Fig. 3, the weather estimation DB 16A registers the weather estimation value for each area in association with each time frame included in each date.

[0043] 3 shows an example of the estimated weather values, which are the temperature and the amount of solar radiation. Note that the weather estimation DB 16A may be data further associated with calendar information. The calendar information may be stored in the storage unit 16 as a database separate from the weather estimation DB 16A.

[0044] Calendar information is information that indicates attributes of each date. The attributes include category information, day of the week information, event information, and the like. Category information is information that indicates whether each day is a weekday, a holiday, or a national holiday, for example. Examples of national holidays include "national holidays" published by the Cabinet Office of Japan, New Year's Day (January 1st to January 3rd), Obon (August 13th to August 16th), and New Year's holidays (December 28th to December 31st). Category information may also include information that indicates time zones such as daytime, late night, and commuting hours. Category information may also include information that indicates seasons such as summer and winter. Day of the week information is information that indicates the day of the week for each day. Event information is information that indicates the name of an event, such as a national holiday, a festival, or a sporting event.

[0045] Returning to Figure 1, we will continue with the explanation.

[0046] The processing information acquisition unit 18A acquires performance data from the performance value management system 3. The processing information acquisition unit 18A transmits an acquisition request for performance data to the performance value management system 3 and acquires performance data from the performance value management system 3. The acquisition request from the performance value management system 3 may include designation information for the area to be acquired. In this case, the performance value management system 3 provides performance data of a point that is the designated area to the information processing device 10. The processing information acquisition unit 18A stores the performance data acquired from the performance value management system 3 in the performance DB of the memory unit 16.

[0047] For example, the processing information acquisition unit 18A may transmit a request to acquire performance data to the performance value management system 3 at regular time intervals, and acquire unacquired performance data from the performance value management system 3. Alternatively, the processing information acquisition unit 18A may specify a target acquisition period, and acquire performance data belonging to the specified period from the performance value management system 3. Alternatively, the processing information acquisition unit 18A may acquire performance data by a push notification from the performance value management system 3. The processing information acquisition unit 18A stores the acquired performance data in a performance DB of the storage unit 16.

[0048] FIG. 4 is a schematic diagram of an example of the data configuration of the actual result DB 16B. The actual result DB 16B is a database in which actual data including the actual value of the electricity market price per unit time of buying and selling for each area in the electricity trading market, such as date, time code, and area. In this example, the actual value of the electricity market price is the electricity market price determined for the area. The electricity market price is expressed in units of 0.01 yen per kWh, but is not limited to this. The time code is expressed by dividing one day into 30 minutes, which is an example of a time frame, and labeling each time frame with an integer value from 1 to 48.

[0049] Returning to Figure 1, we will continue with the explanation.

[0050] The hyperparameters are parameters used in the processing by the processing unit 18, and are parameters that are not determined by the processing by the processing unit 18. The hyperparameters are input by a user's operation instruction on the UI unit 14. In this embodiment, the hyperparameters include a hyperparameter for similarity search, a hyperparameter for persistent estimation, and a hyperparameter for weighting coefficients.

[0051] The hyperparameters for similarity search are hyperparameters used during similarity search.

[0052] The similarity search is a process of using the similarity of weather estimate values ​​to calculate an actual value according to the similarity as an estimate value for a specific time. For example, the similarity search is a process of calculating, as a similarity search estimate value by similarity search, an actual value of a time frame that indicates a weather estimate value according to the similarity with the weather estimate value of the year, month, day, and time frame to be estimated, among the weather estimate values ​​for each of a plurality of times within a certain period of time. The similarity search estimate value is an estimate value calculated by similarity search.

[0053] Specifically, the similarity search is a process of calculating a similarity search estimate value for a future estimated date and time frame of the estimated area from the actual values ​​of the weather estimate values ​​for the estimated date and time frames of the estimated area that are highly similar to the weather estimate values ​​for the estimated date and time frames of the estimated area that are in the future from the specified period of the estimated area, among the weather estimate values ​​for each estimated area for multiple date and time frames included in a specified period (e.g., one month) in the weather estimation DB16A.

[0054] The similarity is calculated by multiplying the two-dimensional space of temperature and solar radiation by the number of meteorological points used, such as a 2N-dimensional space, in which the two-dimensional space is made up of temperature and solar radiation, and the number of meteorological points used is N. The distance is calculated by, for example, Euclidean distance.

[0055] Examples of processes for calculating a similar search estimate from actual values ​​for a year, month, day, and time frame of highly similar weather estimates include processes for calculating the most similar actual value, a simple average of M actual values ​​(M is an integer of 2 or more) in descending order of similarity, or a weighted average obtained by multiplying each of M actual values ​​in descending order of similarity by a weighting value corresponding to each similarity, as the similar search estimate.

[0056] When the electricity market price is the Japanese spot market price, the estimation target area corresponds to one of nine power areas. Therefore, the weather estimate value may use the weather estimate value of multiple points in each area. Assume that the similar search estimate value is calculated for a certain time in the Tokyo area as the power area. In this case, a similar search is possible using the weather estimates of all AMeDAS points included in the Tokyo area, and the user can also specify which point to strongly reference for the similar search.

[0057] In addition, in a similar search, a similar search estimate value for the time to be estimated in the area may be calculated from the actual value of the date and time of the weather estimate value that is highly similar to the weather estimate value for a time in the future of the specified period for the estimated area, among the weather estimate values ​​for each of multiple dates and times that indicate attributes represented by calendar information specified by a user, etc., included in a specified period (e.g., one month) in the estimated area.

[0058] Although details will be described later, the processing unit 18 is capable of calculating multiple types of similarity search estimated values ​​through multiple types of similarity searches by performing calculations using one or multiple types of hyperparameters for similarity search specified by a user's operational instructions, etc.

[0059] Examples of hyperparameters for similarity search include the period used for the similarity search, the start date and time of the period, the attribute of the time used for the similarity search, the value category of the performance value used for the similarity search, the area name used for the similarity search, the type of similarity, a calculation formula for the similarity search estimated value, etc. Examples of the value category of the performance value include, but are not limited to, normal, which is a performance value below a threshold, high, which is a performance value above the threshold or all performance values, etc.

[0060] The hyperparameters for persistent estimation are parameters used for persistent estimation.

[0061] The persistent estimation is a process of calculating an estimated value for a specific time of an estimation target, assuming that a weather estimated value for a time before the specific time of the estimation target continues until the specific time of the estimation target. For example, the persistent estimation is a process of calculating an actual value for a specific time among a plurality of times within a certain period of time, and a moving average of actual values ​​for a specific period within the certain period, as a persistent estimated value by persistent estimation. The persistent estimated value is an estimated value for the time (year, month, day, hour frame) of the estimation target calculated by persistent estimation.

[0062] A method of calculating an actual value of a specific time among a plurality of times within a predetermined period as a persistence estimated value is called simple persistence estimation. The specific time used in simple persistence estimation is preferably the time immediately preceding the specific time to be estimated among the plurality of times within the predetermined period, which is closest to the specific time to be estimated. In addition, when calendar information is used, the specific time used in simple persistence estimation is preferably the time immediately preceding the specific time to be estimated on a day with the same attributes as the day to which the specific time to be estimated belongs, among the plurality of times within the predetermined period, which is closest to the specific time to be estimated.

[0063] The immediately preceding time refers to the time on the previous day if the electricity market price in the electricity trading market is updated daily like the spot market price, and to the time 30 minutes ago (one time frame ago) if the electricity market price in the electricity trading market is updated every 30 minutes or so like the imbalance charge.

[0064] The method of calculating the moving average of the performance values ​​for a specific period within a predetermined period as a sustained estimated value is called moving average estimation. The predetermined period and the specific period are appropriately specified by a user's operation instruction or the like.

[0065] In addition, the simple persistence estimation and the moving average estimation may be used in combination with calendar information. For example, in the simple persistence estimation and the moving average estimation, the persistence estimation value may be calculated in the same manner as described above, using the actual values ​​of the time of the day with the same attribute as the day to which the time of the estimation target belongs among the multiple times in the predetermined period or the specific period. For example, assume that weekday is used as the attribute. In this case, in the case of the simple persistence estimation, the actual value of the time immediately before the time of the estimation target among the multiple times included in the weekday in the predetermined period and in the same time frame as the estimation target is calculated as the persistence estimation value. In addition, in the case of the moving average estimation, the moving average of the actual values ​​of the multiple times included in the weekday in the specific period in the predetermined period and in the same time frame as the estimation target may be calculated as the persistence estimation value.

[0066] Although details will be described later, the processing unit 18 is capable of calculating multiple types of persistence estimate values ​​using multiple types of persistence estimation by calculation using one or multiple types of persistence estimation hyperparameters specified by user operation instructions.

[0067] Examples of hyperparameters for persistence estimation include, but are not limited to, whether to use simple persistence estimation or moving average estimation, whether to use calendar information if it is simple persistence estimation, whether to use calendar information if it is moving average estimation, the start date and time of a predetermined period used for persistence estimation and the end date and time of the predetermined period, the start date and time of a specific period within the predetermined period used for moving average and the end date and time of the specific period, etc. Whether to use calendar information indicates whether to selectively use data for days with specified attributes as data to be used for processing, such as whether to use weekday data or holiday data.

[0068] In addition, the hyperparameters for similarity search and the hyperparameters for sustained estimation may further include information indicating which system provides the weather estimation data and performance data to be used for the similarity search and sustained estimation.

[0069] The weighting coefficient hyperparameter is a hyperparameter used in the correction process described below. The weighting coefficient hyperparameter may be a calculation formula for calculating the weighting coefficient from the deviation. The weighting coefficient hyperparameter, the deviation, and the like will be described in detail later.

[0070] The processing information acquisition unit 18A acquires a first time T1, which is a date, time frame, and time frame to be estimated, in response to an operation instruction by a user on the UI unit 14. In this embodiment, the processing information acquisition unit 18A acquires the first time T1 and the estimation target area. The first time T1 and the estimation target area are inputted in response to an operation instruction by a user on the UI unit 14, and the processing information acquisition unit 18A accepts and acquires the first time T1 and the estimation target area from the UI unit 14.

[0071] Then, when the processing information acquisition unit 18A receives input of the first time T1, which is the year, month, day, and time frame to be estimated, and the estimation target area from the UI unit 14, it acquires, in addition to the received first time T1 and estimation target area, weather estimation data, actual data, and hyperparameters as estimation processing target information to be used in the estimation processing of the estimated value of the received first time T1.

[0072] The processing information acquisition unit 18A acquires the received first time T1 and past weather estimation data up to the first time T1 from the weather estimation DB 16A. In detail, the processing information acquisition unit 18A acquires weather estimation data of the estimation target area for a period including the first time T1 and going back from the first time T1 to a period equal to or longer than the total period of the first predetermined period Q1 and the second predetermined period Q2 from the weather estimation DB 16A. For example, the processing information acquisition unit 18A acquires weather estimation data of the estimation target area for two months including the first time T1 and going back from the first time T1 from the weather estimation DB 16A. At this time, the processing information acquisition unit 18A acquires weather estimation data of at least one point included in the received estimation target area from the weather estimation DB 16A.

[0073] Furthermore, the processing information acquisition unit 18A acquires the latest actual measurement data of the estimation target area up to the received first time T1 from the result DB 16B. In detail, the processing information acquisition unit 18A acquires the actual measurement data of the estimation target area for a period going back from the first time T1 to a period equal to or longer than the total period of the first predetermined period Q1 and the second predetermined period Q2 from the result DB 16B. For example, the processing information acquisition unit 18A acquires the actual measurement data of the estimation target area for two months going back from the first time T1 from the result DB 16B.

[0074] The period of the weather estimation data and the actual data acquired by the processing information acquisition unit 18A from each of the weather estimation DB 16A and the actual DB 16B may be a period according to the extraction condition specified by the user's operation instruction on the UI unit 14. The extraction condition may be included in each of the hyperparameters for similar search and the hyperparameters for maintenance estimation. In addition, an attribute indicating whether weekdays or holidays are used as the extraction condition may be specified. In this case, the processing information acquisition unit 18A may acquire the weather estimation data and the actual data corresponding to only weekdays or only holidays according to the specified attribute during the above period.

[0075] Next, the second estimated value acquisition unit 18B will be described. The second estimated value acquisition unit 18B acquires a second estimated value.

[0076] The second estimated value is an estimated value at the second time T2. As described above, in this embodiment, a case will be described in which the estimated value is an estimated value of the electricity market price per unit time of buying and selling. Furthermore, the second time T2 is a date and time frame within the first specified period Q1 immediately preceding the first time T1, which is the time (date and time frame) to be estimated. In other words, the second estimated value is an estimated value of the electricity market price for a date and time frame within the first specified period Q1 immediately preceding the first time T1, which is the time (date and time frame) to be estimated.

[0077] As described with reference to FIG. 2, the second time T2 is a time within the first predetermined period Q1 going back from the first time T1. The first predetermined period Q1 may be set in advance by a user's operation instruction on the UI unit 14. The first predetermined period Q1 may be included in each of the hyperparameters for similar search and the hyperparameters for maintenance estimation. In this embodiment, the first predetermined period Q1 is assumed to be the seven days immediately preceding the first time T1, which is the date and time frame to be estimated.

[0078] The second estimate value acquisition unit 18B determines whether or not the second estimate value for each of the second times T2 for the seven days immediately preceding the first time T1 is stored in the storage unit 16. The second times T2 for the seven days immediately preceding the first time T1 refer to the second times T2 in the same time frame as the first time T1 for each of the seven days going back in time from the day before the day including the first time T1.

[0079] The second estimate value acquisition unit 18B determines whether a second similarity search estimate value and a second sustained estimate value for the second time T2 of the seven days corresponding to each of the similarity search hyperparameters and sustained estimation hyperparameters included in the estimation processing target information acquired by the processing information acquisition unit 18A are stored in the memory unit 16.

[0080] The second similarity search estimate is an example of a similarity search estimate, and is an estimate of the electricity market price at the second time T2 estimated by similarity search. The second similarity search estimate is a similarity search estimate calculated by one or more types of similarity searches from the second actual value at the second time T2 indicating the second weather estimate according to the similarity with the first weather estimate at the first time T1 among the second weather estimates at each of the multiple second times T2 in the same time frame as the first time T1 within the first predetermined period Q1.

[0081] The second weather estimated value is an example of a weather estimated value, and is a weather estimated value associated with the second time T2 in the weather estimation DB 16A. The first weather estimated value is an example of a weather estimated value, and is a weather estimated value associated with the first time T1 in the weather estimation DB 16A. The second actual value is an example of an actual value, and is a actual value of the electricity market price per unit time of buying and selling at the second time T2. The third actual value is an example of an actual value, and is a actual value of the electricity market price per unit time of buying and selling at the second time T2.

[0082] The second sustained estimate is an example of a sustained estimate, and is an estimate of the electricity market price at the second time T2 calculated by the sustained estimation.

[0083] When the second similar search estimate value and the second sustained estimate value for the second time T2 during the seven days are not stored in the memory unit 16, the second estimate value acquisition unit 18B calculates one or more types of second estimate values, including one or more types of second similar search estimate values ​​and one or more types of second sustained estimate values, for each of the second times T2 during the seven days, using hyperparameters represented by the similarity search hyperparameters and sustained hyperparameters included in the processing target information acquired by the processing information acquisition unit 18A.

[0084] The second estimate value acquisition unit 18B calculates a second similarity search estimate value at the second time T2 by similarity search, and calculates a second maintenance estimate value at the second time T2 by maintenance estimation.

[0085] The similarity search and the maintenance estimation are processes that are also used by functional units other than the second estimated value acquisition unit 18 B. Since the maintenance estimation has been described above, the similarity search will be described in detail.

[0086] -Similar search- The similarity search executed by the processing unit 18 will now be described in detail.

[0087] The processing unit 18 calculates the estimated value by the following formula (1).

[0088] P pred =Model 類似 (DB 過去·気象 ,D 対象·気象 ,hparames 類似 )...Equation (1)

[0089] P pred represents the estimated value of the date, time, and frame to be estimated. pred is an estimate of the electricity market price per unit time of trading, and is calculated for each target date and time frame and target area. 類似 represents a similarity search. DB 過去·気象 is a set of weather estimates for each time frame of the year, month, day, and time prior to the target time frame. D 対象·気象is a set of weather estimates for each of multiple time frames included in the target date. 類似 is a hyperparameter for similarity search.

[0090] In the similarity search, the target date and time frame j = (t_pred,TB pred ) to find weather estimates for past time frames with similar weather conditions. We define a similarity score.

[0091] The similarity score is defined as the distance between real vectors defined below. The processing unit 18 first extracts past weather estimates that belong to the same group of time frames and weekday / holiday classifications as the target date of estimation. The processing unit 18 then standardizes the extracted past weather estimates so that they have a mean of zero and a variance of 1. The processing unit 18 also standardizes the weather estimates for the target date and time frame of estimation using the mean and variance used to standardize the past weather estimates, and defines a real vector expressed by the following formula (2).

[0092]

number

[0093] The similarity score is expressed by the distance of the real vectors (see formula (3)) expressed by formula (2).

[0094] Then, the processing unit 18 calculates the estimated value P by the similarity and a weighted average using the similarity. pred Calculate.

[0095] For example, an example of similarity search parameters used in similarity searches is shown in Table 1. Also, representative examples of AMeDAS points for each power area are shown in Table 2.

[0096] [Table 1]

[0097] [Table 2]

[0098] 1) DB in Eq. (1) 過去·気象 , and D. 対象·気象 About

[0099] A database of weather estimates for each time frame of the year, month, day, and time prior to the target date and time frame for similar searches 過去·気象 If we combine the date t and the time frame TB and organize them by label i as i=(t,TB), we can express it as equation (4).

[0100]

number

[0101] Here, i is a positive integer, days indicates how many past days' worth of weather estimates are to be used as past weather estimates for similarity search, and 48 indicates the number of time frames in a day.

[0102] When performing a similar search using weather estimates for the past 35 days in the Tohoku area, 過去·気象 The number of time frames included is 35 days x 48 hours = 1680.

[0103] In addition, in formula (4), Holiday i is a label that indicates a weekday or a holiday. i T2 indicates the actual value of the electricity market price per unit time for trading in the area. i,n is the meteorological estimate of temperature, AVG4 i,n is a meteorological estimate of the four-grid average solar radiation. The i in these subscripts (i,n) indicates the date and time frame, and n indicates the AMeDAS points included in the area (n=1,...,n code ) If all AMeDAS points in the Tohoku area are used, code =171.

[0104] Similarly, a set D of weather estimates for each of the multiple time frames included in the target date is対象·気象 The estimation target date t and time frame TB are collectively expressed as j=(t pred ,TB pred ) and rearranged by label j, it is expressed by the following equation (5).

[0105]

number

[0106] In formula (5), j is a positive integer, and 48 represents the number of time frames in a day. tpred Since it is common to all of the above, we eliminated the dependency of the subscript and named it Holiday. j,n is the meteorological estimate of temperature, AVG4 j,n is a meteorological estimate of the four-grid average solar radiation. The subscripts (j,n) in these numbers indicate the date and time frame, and n indicates the AMeDAS points included in the area (n=1,...,n code ) If all AMeDAS points in the Tohoku area are used, code =171.

[0107] 2) Extraction of data that matches the target time frame and weekday / holiday labels (corresponding to the parentheses in formula (1))

[0108] The processing unit 18 calculates a set D of weather estimates for each of a plurality of time frames included in the date of estimation. 対象·気象 From the estimation target date and time frame j * Specify one of the following to extract the data, and estimate the weather value d for the target date and time frame. pred Let us assume that d pred can be expressed by the following equation (6).

[0109]

number

[0110] In formula (6), T2 0,n is the meteorological estimate of temperature, AVG4 0,n is a meteorological estimate of the four-grid average solar radiation.

[0111] The processing unit 18 stores a set DB of weather estimates for each time frame of a date in the past from the date and time frame of the estimation target, which is the subject of the similarity search. 過去·気象 From the time frame j to be estimated, * Same time frame as TB * , and the weekday holiday label Holiday that is the same as the target date * The weather estimates for the past date and time frames are extracted. ref,i Let us assume that d ref,i represents the weather estimate for the same time frame as the time frame of the target date among the past weather estimates for the past date and time frames. ref,i can be expressed by the following equation (7).

[0112]

number

[0113] In equation (7), m is the number of data that matches the weekday / holiday condition. I is the time frame of TB. * Since the date is fixed at , only the date has freedom.

[0114] 3) d ref,i Standardization

[0115] The processing unit 18 calculates the past weather estimate d for the same time frame as the time frame to be estimated. ref,i Using these, the temperature and solar radiation (solar radiation intensity) are standardized for each AMeDAS point using the following equations (8) to (13).

[0116]

number

[0117] In formulas (8) to (13), n=1,...,n code is the subscript for each AMeDAS point. T2 ave,n AVG4 represents the average of the climatological estimates of temperature. ave,n represents the average solar radiation of the four grids. T2std,n represents the standard deviation of the climatological estimate of temperature. AVG4 std,n represents the standard deviation of the four-grid solar radiation. Equation (8) represents the meteorological estimate of the standardized temperature. Equation (9) represents the standardized four-grid solar radiation.

[0118] The actual vector of the estimated weather value for the estimation target day shown in the above formula (2) can be defined by the following formula (14).

[0119]

number

[0120] In equation (14), i is the degree of freedom for the date of past weather estimates, i = 1, . . . , m. The real vector represented by equation (14) obtained in this way is 2n code It becomes a +1-dimensional vector.

[0121] 4) d pred Standardization

[0122] Weather estimate for the target date and time frame d pred The average of the meteorological estimates of temperature at each AMeDAS point obtained in 3) above, T2 ave,n , Average of solar radiation for 4 grids AVG4 ave,n , the standard deviation of the meteorological estimate of temperature T2 std,n , Average of solar radiation for 4 grids AVG4 std,n , normalization is performed using the following equations (15) and (16).

[0123]

number

[0124] From the above, the real vector is expressed by the following equation (17).

[0125]

number

[0126] In formula (17), t pred represents the date and time frame to be estimated. The real vector obtained by equation (17) is code It is a +1-dimensional vector.

[0127] 5) Define the distance between the past weather estimates and the weather estimates for the target date and time frame.

[0128] Past weather estimates for the same time frame as the time frame to be estimated d ref,i and the weather estimate d for the target date and time frame pred and the distance U i,c=T2 ,U i,c=AVG4 is expressed by the following equations (18) and (19).

[0129]

number

[0130] The distance between the data for the date i,day are defined by the following equations (20) and (21).

[0131]

number

[0132] Using the above, the distance between the real vector expressed by the above formula (14) and the real vector expressed by the formula (17) is expressed by the following formula (22).

[0133]

number

[0134] In formula (22), w c=T2,AVG4,day are the weights of each element. By default, w c=T2,AVG4,day =1.

[0135] 6) Definition of Similarity

[0136] Past weather estimates for the same time frame as the time frame to be estimated d ref,i and the weather estimate d for the target date and time frame pred The similarity between and is expressed by the inverse of the distance between the above vectors, as shown in the following formula (23).

[0137]

number

[0138] The smaller the distance value expressed by equation (23), the closer the past weather estimate d ref,i and the weather estimate d for the target date and time frame pred The more similar the weather conditions are, the higher the similarity value will be.

[0139] In a similarity search, the degrees of similarity for the dates of past data are calculated for all degrees of freedom i=1,...,m, and then the data are sorted in order of greatest similarity.

[0140] 7) Calculation of similar search estimates based on similar searches on the estimated target date

[0141] For example, the processing unit 18 uses the top K pieces of data with high similarity and calculates a weighted average of the actual values ​​Pi of the electricity market price for each area with the similarity as a weight, thereby calculating a similar search estimated value of the electricity market price for the year, month, day, and time frame to be estimated. K is an integer of 1 or more.

[0142]

number

[0143] Through the above process, the processing unit 18 calculates the similarity search estimated value of the year / month / day / time frame that is the estimation target through similarity search.

[0144] The second estimate value acquisition unit 18B uses each weather estimate value at a third time T3 within the second predetermined period Q2 as a past weather estimate value to be the target of the similarity search. Then, the second estimate value acquisition unit 18B uses the weather estimate value at the third time T3 and the third actual value of the electricity market price to calculate a second similarity search estimate value of each electricity market price at a second time T2 within the first predetermined period Q1 (e.g., the past seven days) going back from the first time T1 to be estimated by the similarity search described above.

[0145] The first predetermined period Q1 may be appropriately changeable by a user's operation instruction via the UI unit 14. The processing unit 18 may obtain, by learning, a period in which the estimation accuracy improves depending on the season or the like as the first predetermined period Q1.

[0146] When the second estimate acquisition unit 18B calculates the second similarity search estimate, P pred represents the weather estimation value at the second time T2, which is the date and time frame to be estimated. As described above, the second time T2 is, for example, the date and time frame of the seven days immediately before the first time T1. In this case, DB 過去·気象 represents a set of weather estimates for a third time T3, which is a date and time frame that is earlier than the second time T2.

[0147] Further, the second estimate value acquisition unit 18B calculates a second sustained estimate value by the sustained estimation. The second sustained estimate value is a sustained estimate value at a second time T2.

[0148] The second estimate value acquisition unit 18B calculates a second sustained estimate value for the second time T2, assuming that the weather estimate value for a time frame prior to the date and time frame to be estimated continues until the second time T2, which is the date and time frame to be estimated.

[0149] It is preferable that the second estimate acquisition unit 18B calculates multiple types of second similar search estimate values ​​based on multiple types of similar searches and multiple types of second sustained estimate values ​​based on multiple types of sustained estimation by adjusting a combination of multiple types of parameters represented by hyperparameters, attributes such as the extraction period and weekdays of past data used for similar search and sustained estimation, a specific time used for sustained estimation, a specific period used for the moving average of sustained estimation, etc.

[0150] For example, it is assumed that the second estimate value acquisition unit 18B prepares two similarity search modes, a normal mode and a price rise mode, and calculates two types of second similarity search estimate values ​​by similarity searches in each mode. It is also assumed that the second estimate value acquisition unit 18B prepares one mode as a sustained estimation mode, and calculates one type of second sustained estimation value by the sustained estimation.

[0151] In normal mode, a database of past weather estimates for similar searches is used. 過去·気象 This means that a set of weather estimates for past date and time frames whose actual values ​​for each date and time frame are below a certain value is used as the "high-rise mode." The high-rise mode uses a set DB of weather estimates for past date and time frames used for similarity search. 過去·気象 This means that, among the weather estimates for past date and time frames, a set of weather estimates whose actual values ​​for each date and time frame are equal to or greater than a predetermined value is used, or the weather estimates for past date and time frames are used regardless of the actual values.

[0152] Specifically, the case where the values ​​shown in Table 3 are used as the hyperparameters for similarity search in the normal mode of similarity search and the parameters for similarity search in the price rise mode of similarity search will be described as an example. In addition, the normal mode DB shown in Table 4 is used as the weather estimated value and the third actual value of the electricity market price at the third time T3, which is a past time, used in the normal mode. 通常 and high-rise mode DB 高騰 As an example, a case will be described in which data represented by each of the following is used.

[0153] [Table 3]

[0154] [Table 4]

[0155] In the similarity search, there is a degree of freedom in the database of weather estimates and actual values, etc., used, and the hyperparameters for similarity calculation. Here, a case will be described in which two types of second similarity search estimates are calculated by performing two types of similarity searches using a characteristic database that is mainly used. In this case, the second estimate value acquisition unit 18B calculates two types of second similarity search estimates for each of the second times T2 in the seven days immediately preceding the first time T1, by the above-mentioned calculation method, using the database shown in Table 4 and the hyperparameters for similarity calculation shown in Table 3. The second similarity search estimate calculated by the similarity search in the normal mode is P1, and the second similarity search estimate calculated by the similarity search in the price rise mode is P2.

[0156] The second estimate acquisition unit 18B calculates a second sustained estimate for each second time T2 in the seven days immediately preceding the first time T1 by sustained estimation that distinguishes between weekdays and holidays. This second sustained estimate is P3. The sustained estimation that distinguishes between weekdays and holidays is a method in which the same weekday or holiday as the target date of estimation on another date closest to the target date of estimation, and the actual value of the same time frame as the target time frame is used as the estimated value of the target date and time frame. The sustained estimation that distinguishes between weekdays and holidays is a method in which the moving average value of the group of weekdays or holidays that are the same as the target date of estimation and the group of actual values ​​of the same time frame as the target time frame in the period of the past few days closest to the target date of estimation is used as the estimated value of the target date and time frame.

[0157] The second estimate acquisition unit 18B may calculate three or more types of second similarity search estimates by three or more types of similarity searches in which at least one of the similarity search hyperparameters and the database is different.

[0158] For example, the second estimate value acquisition unit 18B may calculate two more types of second similar search estimate values ​​by similarity search by using a database of weather estimation data provided by the Japan Weather Association and a database of weather estimation data provided by the National Weather Service as sets of past weather estimation values. Also, the second estimate value acquisition unit 18B may calculate two more types of second similar search estimate values ​​by similarity search by using a database of weather estimation data provided by WRF and a database of weather estimation data provided by MSM (Meso Scale Model).

[0159] Further, the second estimate value acquisition unit 18B may calculate two or more types of second persistent estimate values ​​by two or more types of persistent estimation in which at least one of the persistent estimation hyperparameters and the database is different.

[0160] By the processing unit 18 using a plurality of types of similarity searches and a plurality of types of persistent estimation, it becomes possible to calculate an estimate of the first time T1 with higher accuracy.

[0161] Through these processes, when the second estimate values ​​for each of the second times T2 for the seven days immediately preceding the first time T1 are stored in the memory unit 16, the second estimate value acquisition unit 18B reads the second estimate values ​​from the memory unit 16 to acquire the second estimate values ​​for each of the second times T2 for the seven days immediately preceding the first time T1.

[0162] In addition, if the second estimate values ​​for each of the second times T2 in the seven days immediately preceding the first time T1 are not stored in the memory unit 16, the second estimate value acquisition unit 18B calculates the second estimate values ​​for each of the second times T2 in the seven days immediately preceding the first time T1 by the above calculation process, and acquires the second estimate values.

[0163] We will assume a situation in which the second estimate acquisition unit 18B acquires, for each of the second times T2 in the past seven days immediately prior to the first time T1, a second similar search estimate P1 calculated by a similar search in normal mode, a second similar search estimate P2 calculated by a similar search in price increase mode, and a second sustained estimate P3 calculated by sustained estimation.

[0164] The deviation calculation unit 18C calculates the deviation between the second estimated value and the second actual value at a second time T2 within a first predetermined period Q1 going back from the first time T1, which is the date and time frame to be estimated.

[0165] In detail, the deviation calculation unit 18C calculates the deviation for each estimation method i. The estimation method i represents a type (method) of similar search and a type (method) of persistent estimation.

[0166] That is, the deviation calculation unit 18C calculates deviations including similarity search deviations between the second estimated value, which is a second similarity search estimated value calculated by one or more types of similarity search from the second actual value at the second time T2 indicating the second weather estimated value according to the similarity with the first weather estimated value at the first time T1, among the second weather estimated values ​​at each of the multiple second times T2 in the first predetermined period Q1, and the second actual value. The similarity search deviations represent the deviation between the second similarity search estimated value at the second time T2 calculated by similarity search and the second actual value at the second time T2.

[0167] Moreover, the deviation calculation unit 18C calculates deviations including a deviation between the second actual value and the second sustained estimated value calculated by one or more types of sustained estimation including at least one of the second actual value at a specific second time T2 among the multiple second times T2 in the first predetermined period Q1 and a moving average of the second actual value for a specific period in the first predetermined period Q1, and a similar search deviation. The sustained deviation represents a deviation between the second sustained estimated value at the second time T2 calculated by the sustained estimation and the second actual value at the second time T2.

[0168] For example, we will assume a situation in which the second estimate acquisition unit 18B acquires a second similar search estimate value P1 calculated by a similar search in normal mode, a second similar search estimate value P2 calculated by a similar search in price increase mode, and a second sustained estimate value P3 calculated by sustained estimation for each second time T2 that is the same time frame as the first time T1, which is the date and time frame to be estimated, during the past seven days immediately prior to the first time T1.

[0169] In this case, the deviation calculation unit 18C calculates the deviation for each estimation method i between the second similarity search estimated value P1[t,i] calculated by the similarity search in the normal mode, the second similarity search estimated value P2[t,i] calculated by the similarity search in the price rise mode, and the second sustained estimated value P3[t,i] calculated by the sustained estimation, and the second actual value P0 of the electricity market price at the second time T2, for each second time T2 in the past seven days immediately before the first time T1. In this case, the deviation calculation unit 18C calculates three types of deviation for each estimation method i.

[0170] When the deviation calculation unit 18C calculates the deviation between one type of second estimated value and the second actual value for the first time T1 to be estimated, the error (difference) between the second estimated value and the second actual value is used as the deviation. When the deviation calculation unit 18C calculates the deviation between each of multiple types of second estimated values ​​and the second actual value for the first time T1 to be estimated, the mean absolute error (MAE) or root mean square error of these multiple types of deviations is used as the deviation. For example, a case where the MAE (mean absolute error) is used as the deviation will be described.

[0171]

number

[0172] In equation (25), MAE represents the deviation. P j [t, i] represents a second estimated value calculated by the estimation method i at the time frame t, which is the second time T2. P0[t, i] represents an actual value at the time t, which is the second time T2.

[0173] Next, the estimation unit 18D will be described.

[0174] The estimation unit 18D calculates a first estimated value of the first time T1 according to third actual values ​​of a plurality of third times T3 within a second predetermined period Q2 that includes a period before the first time T1 and the second time T2.

[0175] For example, the estimation unit 18D calculates, as the first estimate, a first similar search estimate value calculated by one or more types of similarity search from a third actual value at the third time T3 indicating a third weather estimate value according to the similarity with the first weather estimate value at the first time T1 among the third weather estimate values ​​at each of the multiple third times T3. The third weather estimate value is a weather estimate value at the third time T3. The third actual value represents an actual value at the third time T3. The first similar search estimate value is an estimate value at the first time T1 estimated by similarity search.

[0176] In addition, the estimation unit 18D may calculate, as the first estimate value, a first sustained estimate value calculated by one or more types of sustained estimation including at least one of a third actual value among multiple third times T3 and a moving average of the third actual values ​​for a specific period within the second specified period Q2, and a first similar search estimate value.

[0177] The estimation unit 18D uses the same hyperparameters as the second estimate value acquisition unit 18B, except that it uses the first time T1 instead of the second time T2 as the estimation target, and calculates a first estimate value of the first time T1 using the same estimation method i as the second estimate value acquisition unit 18B, in the same manner as the second estimate value acquisition unit 18B.

[0178] It is assumed that the second estimate acquisition unit 18B acquires, for each second time T2 in the previous seven days, a second similar search estimate P1[t,i] calculated by a similar search in normal mode, a second similar search estimate P2[t,i] calculated by a similar search in price-rise mode, and a second sustained estimate P3[t,i] calculated by sustained estimation.

[0179] In this case, the estimation unit 18D uses the weather estimate value and actual value for the third time T3 within the second specified period Q2 to calculate a first similar search estimate value calculated by similar search in normal mode, a first similar search estimate value calculated by similar search in high price mode, and a first sustained estimate value calculated by sustained estimation as a first estimate value for the first time T1.

[0180] The correction unit 18E corrects the first estimated value according to the deviation. That is, the correction unit 18E calculates a corrected value obtained by correcting the first estimated value at the first time T1 of the estimation target estimated by the estimation unit 18D using the deviation, as an estimated value of the electricity market price at the first time T1.

[0181] In detail, the correction unit 18E corrects the first estimated value by a weighting coefficient according to the deviation. The weighting coefficient is expressed by a function in which the value of the weighting coefficient increases as the deviation approaches zero. Examples of the weighting coefficient expressed by such a function include the reciprocal of the square of the deviation, the reciprocal of the deviation, etc.

[0182] In detail, when the error between one type of second estimated value and the second actual value is used as the deviation, the inverse of the square of the deviation is used as the weighting coefficient represented by the function, for example. When the mean absolute error (MAE) is used as the deviation, the inverse of the square of the deviation is used as the weighting coefficient represented by the function, for example. When the root mean square error is used as the deviation, the inverse of the deviation is used as the weighting coefficient. The combination of these functions used for the deviation and the weighting coefficient can be appropriately changed according to the user's operation instruction of the UI unit 14. That is, the combination of these functions is input as a hyperparameter for the weighting coefficient by the user's operation instruction of the UI unit 14.

[0183] Here, a case will be described in which the mean absolute error (MAE) is used as the deviation and the reciprocal of the square of the deviation is used as the weighting coefficient.

[0184] The correction unit 18E uses a weighting coefficient to correct the first estimation value to a weighted average value obtained by weighting and averaging multiple types of first estimation values ​​for the first time T1 calculated by the estimation unit 18D (for example, a first similar search estimation value calculated by a similar search in normal mode, a first similar search estimation value calculated by a similar search in high price mode, and a first sustained estimation value calculated by sustained estimation, etc.).

[0185] Specifically, the correction unit 18E calculates the weighting coefficient using the following formula (26).

[0186]

number

[0187] W[i] represents a weighting coefficient. w1[i] represents the weight for the deviation between the second similarity search estimated value P1[t,i] calculated by the similarity search in normal mode and the second actual value. w2[i] represents the weight for the deviation between the second similarity search estimated value P2[t,i] calculated by the similarity search in high price mode and the second actual value. w3[i] represents the weight for the deviation between the second maintenance estimated value P3[t,i] calculated by the maintenance estimation and the second actual value. w j [i] represents the degree of deviation between the second estimated value calculated by the estimation method j and the second actual value. The estimation method j here represents the similar search in the normal mode, the similar search in the price increase mode, and the maintenance estimation.

[0188] (W,w j ) represents the weight of the weighted sum. The data for the last 7 days before the first time T1 is -1,...,-7 It is expressed as:

[0189] For time frame [i], the weights for each estimation method are w j = 1, 2, 3, the first estimate of estimation method j is P j , the deviation MAE for the last 7 days is MAE(P j [i]). The first estimate of the electricity market price at the first time T1 is expressed as P jThe correction unit 18E uses the weights thus obtained to calculate the weighted average result P mixed [t0, i] is obtained as a corrected estimate of the first time T1.

[0190]

number

[0191] That is, when the correction unit 18E uses the mean absolute error (MAE) as the deviation, it uses the inverse of the square of the deviation as the weighting coefficient, and calculates a weighted sum with the inverse of the sum of the squares of the deviation as the weighting, as the corrected estimate value for the first time T1.

[0192] In addition, when the correction unit 18E uses the root mean square error as the deviation, the inverse of the deviation is used as the weighting coefficient, and the weighted sum with the inverse of the deviation as the weight is calculated as the corrected estimate value at the first time T1.

[0193] The estimated value of the first time T1 calculated by the correction unit 18E, i.e., the corrected estimated value obtained by correcting the first estimated value of the first time T1 calculated by the estimation unit 18D according to the deviation degree, can be expressed by the following equation (28).

[0194]

number

[0195] In formula (28), D 電力市場価格·推定値 represents the estimated value of the electricity market price for each area at the first time T1. pred is a first estimated value of the first time T1 for each area calculated by the estimation unit 18D. T represents the date and time frame that is the first time T1. A represents the area of ​​the power.

[0196] As a final result of the above processing by the processing unit 18, an estimated value (yen / kWh) per unit time of buying and selling for the date and time frame of the estimation target is obtained for each date and time frame of the estimation target and for each estimation target area.

[0197] The output control unit 18F outputs the estimated value of the electricity market price at the first time T1 calculated by the correction process by the correction unit 18E to at least one of the UI unit 14 and an external information processing device connected via the network NW.

[0198] Fig. 5 is an explanatory diagram of an example of output information by the output control unit 18F. Fig. 5 shows, as an example of output information, the calculation result of the processing unit 18 calculating the estimated value of the electricity market price in area A for each of the first times T1, which are each time frame on September 9, 2022. In this way, by the above processing by the processing unit 18, as a final result, an estimated value (yen / kWh) per unit time of buying and selling to be estimated is obtained for each year / month / day / time frame to be estimated and for each area to be estimated.

[0199] Next, an example of the flow of information processing executed by the processing unit 18 of the information processing device 10 will be described.

[0200] FIG. 6 is a flowchart showing an example of the flow of information processing executed by the processing unit 18.

[0201] The processing information acquisition unit 18A acquires estimation processing target information (step S100). When the processing information acquisition unit 18A receives an input of the first time T1, which is the year, month, day, and time frame of the estimation target, and the estimation target area from the UI unit 14, the processing information acquisition unit 18A reads the received first time T1 and the estimation target area as well as the weather estimation data, the performance data, and the hyperparameters from the storage unit 16 as estimation processing target information used in the estimation processing of the estimated value of the received first time T1. Through these processes, the processing information acquisition unit 18A acquires the estimation processing target information.

[0202] The second estimate value acquisition unit 18B determines whether or not the second estimate value of each second time T2 within the first predetermined period Q1 going back from the first time T1 is stored in the storage unit 16 (step S102). For example, the second estimate value acquisition unit 18B determines whether or not the second estimate value of the second time T2 of the same time frame as the first time T1 is stored in the storage unit 16 during the seven days immediately prior to the first time T1, which is the date and time frame to be estimated acquired in step S100.

[0203] If the answer is affirmative in step S102 (step S102: Yes), the process proceeds to step S106. If the answer is negative in step S102 (step S102: No), the process proceeds to step S104.

[0204] In step S104, the second estimate acquisition unit 18B calculates a plurality of types of second estimates including one or more types of second similarity search estimates and one or more types of second maintenance estimates for each second time T2 in the seven days immediately preceding the first time T1, which is the date and time frame to be estimated, using hyperparameters represented by the similarity search hyperparameters and the maintenance hyperparameters included in the processing target information acquired in step S100.Then, the process proceeds to step S106.

[0205] In step S106, the deviation calculation unit 18C calculates the deviation between the second estimated value for the second time T2 acquired by the second estimated value acquisition unit 18B in step S102 or step S104 and the second actual value for the second time T2 (step S106).

[0206] The estimation unit 18D calculates a first estimate value of the first time T1 according to third actual values ​​of a plurality of third times T3 within a second predetermined period Q2 including a period before the first time T1 to be estimated and the second time T2 immediately before the first time T1 (step S108). As described above, the estimation unit 18D calculates, as the first estimate value, for example, a first similarity search estimate value calculated by similarity search and a first sustained estimate value calculated by sustained estimation. That is, the estimation unit 18D uses the same hyperparameters as the second estimate value acquisition unit 18B, except that the first time T1 is used instead of the second time T2 as the estimation target, and calculates the first estimate value of the first time T1 by the same estimation method i as the second estimate value acquisition unit 18B in the same manner as the second estimate value acquisition unit 18B.

[0207] Assume that the second estimate acquisition unit 18B acquires a second similarity search estimate P1 calculated by similarity search in normal mode, a second similarity search estimate P2 calculated by similarity search in high-price mode, and a second sustained estimate P3 calculated by sustained estimation for each second time T2 in the previous seven days. In this case, the estimation unit 18D calculates the first similarity search estimate P1 calculated by similarity search in normal mode, the first similarity search estimate P2 calculated by similarity search in high-price mode, and the first sustained estimate P3 calculated by sustained estimation as the first estimate at the first time T1, using the weather estimate and actual values ​​at the third time T3 within the second predetermined period Q2.

[0208] The correction unit 18E corrects the first estimated value calculated in step S108 according to the deviation calculated in step S106 (step S110). That is, the correction unit 18E corrects the first estimated value at the first time T1, which is the date and time frame of the estimation target estimated by the estimation unit 18D, using the deviation, and calculates the corrected value as the estimated value of the electricity market price at the first time T1.

[0209] The output control unit 18F outputs the estimated value of the electricity market price at the first time T1 calculated by the correction process in step S110 to at least one of the UI unit 14 and an external information processing device connected via the network NW (step S112). Then, this routine ends.

[0210] As described above, the processing unit 18 of the information processing device 10 of this embodiment calculates a first estimated value for the first time T1 based on third actual values ​​for multiple third times T3 within a second predetermined period Q2 that includes a period prior to a second time T2 within a first predetermined period Q1 going back from the first time T1 to be estimated, and corrects the first estimated value based on the degree of deviation between the second estimated value and the second actual value for the second time T2.

[0211] In the conventional technology, the estimation was performed using the average of actual values ​​of multiple past days with similar weather conditions. Therefore, in the conventional technology, it was not possible to respond to fluctuations in actual values ​​in periods closer to the date and time frame of the estimation target, such as immediately before the estimation target date, and the estimation accuracy may have decreased. In detail, in the conventional technology, the estimated value calculated by a similarity search that extracts the average behavior of actual values ​​for about the past month was used as the estimated value for the date and time frame of the estimation target. Therefore, in the conventional technology, the most recent actual value was not given importance, and it took time for the most recent fluctuations in the actual value to be reflected as past data, and the estimation accuracy during that time may have deteriorated. In addition, in the conventional technology, it was not possible to fully estimate price increases and decreases over a short span, and the estimation accuracy may have deteriorated. In other words, in the conventional technology, it was difficult to accurately estimate the estimated value at the first time T1 of the estimation target.

[0212] On the other hand, the processing unit 18 of the information processing device 10 of this embodiment corrects the first estimated value of the first time T1 calculated according to the third actual values ​​of multiple third times T3 within a second predetermined period Q2 including the period prior to the first time T1 and the second time T2, using the deviation between the second estimated value and the second actual value of the second time T2 within a first predetermined period Q1 going back from the first time T1 to be estimated.

[0213] Therefore, in the information processing device 10 of this embodiment, it is possible to respond to fluctuations in actual values ​​closer to the first time T1 to be estimated, such as immediately before the first time T1 to be estimated, and to suppress a decrease in estimation accuracy.

[0214] Therefore, the information processing device 10 of this embodiment can suppress a decrease in estimation accuracy of the estimation value of the estimation target.

[0215] Moreover, the information processing device 10 of the present embodiment can estimate an estimated value at the first time T1 with improved follow-up to trend changes in the performance values ​​for the period immediately prior to the first time T1 to be estimated.

[0216] That is, in the information processing device 10 of the present embodiment, the degree of deviation from the estimated value at the time when the actual value is obtained can be calculated, and the degree of deviation can be used to correct the latest first estimated value. By appropriately correcting the first estimated value using the latest actual value, it is possible to improve the trackability of the estimated value at the first time T1 to the irregular behavior of the actual value. Therefore, in the information processing device 10 of the present embodiment, it is possible to estimate the estimated value at the first time T1 that tracks the fluctuation of the actual value on a short time scale of several days after the fact.

[0217] In the information processing device 10 of the present embodiment, the processing unit 18 calculates, as a first estimated value, a first similarity search estimated value calculated by one or more types of similarity search from a third actual value at a third time indicating a third weather estimated value according to the similarity with the first weather estimated value at the first time T1 among the third weather estimated values ​​at each of a plurality of third times T3. Then, the processing unit 18 corrects the first estimated value by using the deviation degree.

[0218] Therefore, in the information processing device 10 of this embodiment, the first estimated value, which is an estimation result by similarity search, can be corrected with the most recent actual value at the first time T1, and an estimated value at the first time T1 with improved estimation accuracy can be calculated. Therefore, in the information processing device 10 of this embodiment, it becomes possible to calculate an estimated value at the first time T1 with high accuracy even for the first time T1 in a period in which the trend of actual values ​​is unstable, which makes similarity search difficult.

[0219] In addition, in the information processing device 10 of the present embodiment, a first sustained estimate value calculated by one or more types of sustained estimates including at least one of a third actual result value at any of a plurality of third times T3 and a moving average of the third actual result values ​​for a specific period within the second predetermined period Q2, and a first similar search estimate value are calculated as the first estimate value. Then, the processing unit 18 corrects the first estimate value by using the deviation degree.

[0220] For example, because the spot market price, which is one of the targets of the estimation, is affected by weather estimates, historical data from the past one to two months is required to obtain an average estimate. Empirical evidence has shown that this average estimate has good accuracy during periods when the weather is calm, such as spring and autumn, and when there is no tight supply and demand. On the other hand, during periods when actual values ​​fluctuate greatly, such as when prices rise in winter and summer, average estimates based on weather estimates alone do not provide sufficient accuracy. In extreme cases, persistent estimation, in which the actual value of the electricity market price from the previous day is used as the estimate for the next day, may be better.

[0221] Therefore, by using both the first similarity search estimate and the first sustained estimate as the first estimate and further correcting the first estimate using the deviation, it is possible to calculate an estimate of the first time T1 with higher accuracy.

[0222] <Example of operation> Next, a specific example of the operation of the information processing device 10 of the present embodiment will be described.

[0223] Assume that the similarity search and persistence estimation used by the processing unit 18 are as shown in Tables 5 and 6 below. Here, the number of days of past data is changed from the data in the specific example described in the above embodiment. For persistence estimation, a method is used that distinguishes between weekdays and holidays, and references the actual values ​​for the same time frame on the nearest weekday or holiday to obtain an estimated value.

[0224] [Table 5]

[0225] [Table 6]

[0226] The processing executed by the processing unit 18 of the information processing device 10 of this embodiment will be described below.

[0227] The date and time frame that is the first time T1 to be estimated is 2020 / 09 / 09 10:00~10:30. Figure 8 shows the data in Figure 7, which contains the weather estimates and actual results entered into a single CSV file. If you extract only the data for the 10:00 time frame, which is the same time frame as the first time T1 to be estimated, from the data in Figure 7, you will get the data in Figure 8.

[0228] FIG. 8 is a diagram showing a set of estimated weather data for a time frame prior to the first time T1, and estimated weather data for the time frame of the first time T1 that is the estimation target.

[0229] In this case, the weather estimation data used by the processing unit 18 as the processing target information is expressed by the following formula (29).

[0230]

number

[0231] In equation (29), JST represents the date and time frame, areaPrice2 represents the actual value of the electricity market price in the Tohoku area, T2_34392 represents the temperature (℃) at 2 m above ground at the AMeDAS point in Sendai, and AVG4_34392 represents the average solar radiation (W / m 2 ) Sendai is one of the locations included in the Tohoku area. Holiday is the label of the weekday or holiday represented by the calendar information, where 0 indicates a weekday and 1 indicates a holiday.

[0232] The processing unit 18 executes the similarity search described in the above embodiment using the data shown in Fig. 8. The processing unit 18 standardizes the temperature and the amount of solar radiation with an average of zero and a variance of 1, respectively, and calculates the distance from the estimated value of the weather for the year, month, day, and time frame to be estimated. The calculated distance is represented by SCORE, and is expressed by the following formulas (30) and (31).

[0233]

number

[0234] SCORE K indicates the distance of the weather estimate, and SCORE T indicates the distance in dates. The distance in dates is expressed as the time difference between dates and times.

[0235] Then, the processing unit 18 calculates the similarity from the distance. In detail, the processing unit 18 calculates the similarity Weight from the distance SCORE by the following formula (33).

[0236]

number

[0237] The label i in formula (33) represents a numerical value from 1 to K, and indicates the number of searches used in the similarity search.

[0238] Fig. 9 shows the calculation results calculated by the processing unit 18 through a similarity search in normal mode from the data shown in Fig. 8. Fig. 10 shows the calculation results calculated by the processing unit 18 through a similarity search in price rise mode from the data shown in Fig. 8.

[0239] The processing unit 18 calculates the estimated value of the similarity search by using the similarity Weight and the actual value P of the electricity market price according to the following formula (34).

[0240]

number

[0241] In equation (34), mode represents the similarity search mode, such as normal mode or high-priority mode.

[0242] Furthermore, the processing unit 18 refers to the time frame (2022 / 09 / 08 10:00) of the closest day within the same weekday as the year / month / day time frame that is the first time T1 to be estimated, as the first estimated value of the persistence estimation.

[0243] Through the above process, the processing unit 18 calculates the second estimated value of each electricity market price at the second time T2, which is the same time frame as the first time T1 in the past four days immediately before the first time T1, as shown in Fig. 11. Fig. 11 also shows the second actual value of each electricity market price at each time frame.

[0244] Furthermore, the processing unit 18 calculates a first estimated value for 2022 / 09 / 09 10:00 as the first time T1, which is the year, month, day, and time frame to be estimated. The calculation result is shown in FIG.

[0245] Then, the processing unit 18 calculates the degree of deviation between the second estimated value and the second actual value. Here, the deviation is taken as MAE (mean absolute error), and the degree of deviation between the second estimated value and the second actual value for the past four days immediately prior to the first time T1 is calculated. Then, the processing unit 18 uses this deviation to calculate a weighting coefficient. The weighting coefficient is the reciprocal of the deviation MAE squared.

[0246] The processing unit 18 calculates a weighting coefficient w for each estimation method i. i is calculated by the following formula (35): The estimation method i represents the type (method) of similarity search and the type (method) of persistent estimation.

[0247]

number

[0248] 7 in formula (35), mae i represents the average deviation of estimation method i and is calculated by the following formula (36).

[0249]

number

[0250] In formula (36), mae i,j represents the deviation of estimation method i for date j.

[0251] Then, the processing unit 18 calculates a weighting coefficient w i Using the above, a weighted average of the first estimated values ​​at the first time T1 of the estimation target calculated by each of the two types of similarity searches and one type of persistent estimation is calculated by the following equation (37).

[0252]

number

[0253] The calculation results are shown in FIG. 13 and FIG.

[0254] The results of the processing unit 18 executing these processes for each of the 48 time frames included in the estimation target 2022 / 09 / 09 are shown in Fig. 15. As shown in Fig. 15, it was confirmed that the corrected estimated value, which is the value obtained by correcting the first estimated value based on the deviation, is closer to the actual value of the electricity market price for the corresponding time frame than the first similarity search estimated value obtained by the similarity search in the normal mode.

[0255] Therefore, it was confirmed that the processing unit 18 of the information processing device 10 of this embodiment can suppress a decrease in the estimation accuracy of the estimation value of the estimation target.

[0256] Next, an example of the hardware configuration of the information processing device 10 of the above embodiment will be described.

[0257] FIG. 16 is a diagram showing an example of a hardware configuration of the information processing device 10 according to the embodiment.

[0258] The information processing device 10 of the above embodiment has a hardware configuration utilizing a conventional computer, in which a CPU (Central Processing Unit) 81, a ROM (Read Only Memory) 82, a RAM (Random Access Memory) 83, a communication I / F 84, etc. are interconnected via a bus 85.

[0259] The CPU 81 is a calculation device that controls the information processing device 10 of the above embodiment. The ROM 82 stores programs and the like that realize various processes by the CPU 81. Although a CPU is used for the explanation here, a GPU (Graphics Processing Unit) may be used as the calculation device that controls the information processing device 10. The RAM 83 stores data necessary for various processes by the CPU 81. The communication I / F 84 is an interface for transmitting and receiving data.

[0260] In the information processing device 10 of the above embodiment, the CPU 81 reads out a program from the ROM 82 onto the RAM 83 and executes it, thereby realizing each of the above functions on the computer.

[0261] The programs for executing the above-mentioned processes executed by the information processing device 10 of the embodiment may be stored in a HDD (hard disk drive). Also, the programs for executing the above-mentioned processes executed by the information processing device 10 of the embodiment may be provided by being pre-installed in the ROM 82.

[0262] The program for executing the above-mentioned processes executed by the information processing device 10 of the above-mentioned embodiment may be stored in a computer-readable storage medium such as a CD-ROM, CD-R, memory card, DVD (Digital Versatile Disk), or flexible disk (FD) in an installable or executable format file and provided as a computer program product. The program for executing the above-mentioned processes executed by the information processing device 10 of the above-mentioned embodiment may be stored on a computer connected to a network such as the Internet and provided by downloading the program via the network. The program for executing the above-mentioned processes executed by the information processing device 10 to information processing device 10E of the above-mentioned embodiment may be provided or distributed via a network such as the Internet.

[0263] Although the embodiment of the present invention has been described above, this embodiment is presented as an example and is not intended to limit the scope of the invention. This new embodiment can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the gist of the invention. This embodiment and its modifications are included in the scope and gist of the invention, and are included in the scope of the invention and its equivalents described in the claims.

[0264] The present technology can also be configured as follows. (1) calculating a first estimated value of the first time according to third actual values ​​at a plurality of third time points within a second predetermined period including a period prior to a second time point within a first predetermined period going back from the first time point to be estimated; a processing unit that corrects the first estimated value in accordance with a degree of deviation between a second estimated value at the second time and a second actual value; An information processing device comprising: (2) The processing unit includes: Among the third weather estimated values ​​at the plurality of third times, a first similarity search estimated value calculated by one or more types of similarity search from a third actual value at the third time indicating the third weather estimated value according to the similarity with the first weather estimated value at the first time is calculated as the first estimated value; An information processing device as described in (1). (3) The processing unit includes: calculating, as the first estimated value, a first sustained estimated value calculated by one or more types of sustained estimation including at least one of the third actual value at any of the plurality of third time points and a moving average of the third actual value for a specific period within the second predetermined period, and the first similar search estimated value; An information processing device according to (2). (4) The processing unit includes: Calculating the deviation degree including the similarity search deviation degree between the second estimated value, which is a second similarity search estimated value calculated by one or more types of similarity searches from a second actual value at the second time that indicates the second weather estimated value according to the similarity to the first weather estimated value at the first time, among the second weather estimated values ​​at each of the plurality of second times within the first predetermined period, and the second actual value; An information processing device according to any one of (1) to (3). (5) The processing unit includes: calculating a degree of deviation between a second sustained estimated value calculated by one or more types of sustained estimation including at least one of a second performance value at a specific second time among the plurality of second times within the first predetermined period and a moving average of the second performance values ​​for a specific period within the first predetermined period, and the second performance value, and the degree of deviation including the similar search deviation; (4) An information processing device. (6) The processing unit includes: correcting the first estimated value by a weighting coefficient according to the degree of deviation; An information processing device according to any one of (1) to (5). (7) The weighting coefficients are: The weighting coefficient is expressed by a function in which the value of the weighting coefficient increases as the value of the deviation approaches zero. The reciprocal of the square of the deviation, or the reciprocal of the deviation, (6) An information processing device. (8) the first estimated value, the second estimated value, and the third estimated value at the third time are estimated values ​​of an electricity market price per unit time of buying and selling in an electricity trading market; The second actual value and the third actual value are actual values ​​of the electricity market price per unit time of the buying and selling. An information processing device according to any one of (1) to (7). (9) An information processing method executed by an information processing device, calculating a first estimated value of the first time according to third actual values ​​at a plurality of third time points within a second predetermined period including a period prior to a second time point within a first predetermined period going back from the first time point to be estimated; correcting the first estimated value according to a degree of deviation between the second estimated value at the second time and a second actual value; Information processing methods. (10) On the computer, calculating a first estimated value of the first time according to third actual values ​​of a plurality of third time points within a second predetermined period including a period prior to a second time point within a first predetermined period going back from the first time point to be estimated; correcting the first estimated value according to a degree of deviation between the second estimated value at the second time and a second actual value; Information processing program. [Explanation of symbols]

[0265] 1. Information Processing Systems 10. Information processing device 18 Processing section 18A Processing information acquisition unit 18B Second estimated value acquisition section 18C Deviation calculation section 18D Estimation Section 18E Correction section

Claims

1. calculating a first estimated value of the first time according to third actual values ​​of a plurality of third time points within a second predetermined period including a period prior to the second time point within a first predetermined period going back from the first time point to be estimated; a processing unit that corrects the first estimated value in accordance with a degree of deviation between a second estimated value at the second time and a second actual value; An information processing device comprising:

2. The processing unit includes: Among the third weather estimated values ​​at the plurality of third times, a first similarity search estimated value is calculated by one or more types of similarity search from a third actual value at the third time indicating the third weather estimated value according to the similarity with the first weather estimated value at the first time, and is calculated as the first estimated value. The information processing device according to claim 1 .

3. The processing unit includes: a first sustained estimation value calculated by one or more types of sustained estimation including at least one of the third actual value at any of the plurality of third time points and a moving average of the third actual value for a specific period within the second predetermined period, and the first similar search estimation value are calculated as the first estimated value; The information processing device according to claim 2 .

4. The processing unit includes: Calculating the deviation degree including the similarity search deviation degree between the second estimated value, which is a second similarity search estimated value calculated by one or more types of similarity searches from a second actual value at the second time that indicates the second weather estimated value according to the similarity to the first weather estimated value at the first time, among the second weather estimated values ​​at each of the plurality of second times within the first predetermined period, and the second actual value; The information processing device according to claim 1 .

5. The processing unit includes: calculating a degree of deviation between a second sustained estimated value calculated by one or more types of sustained estimation including at least one of a second performance value at a specific second time among the plurality of second times within the first predetermined period and a moving average of the second performance values ​​for a specific period within the first predetermined period, and the second performance value, and the degree of deviation including the similar search deviation; The information processing device according to claim 4.

6. The processing unit includes: correcting the first estimated value by a weighting coefficient according to the degree of deviation; The information processing device according to claim 1 .

7. The weighting coefficients are The weighting coefficient is expressed by a function in which the value of the weighting coefficient increases as the value of the deviation approaches zero. The information processing device according to claim 6.

8. the first estimated value, the second estimated value, and the third estimated value at the third time are estimated values ​​of an electricity market price per unit time of buying and selling in an electricity trading market, The second actual value and the third actual value are actual values ​​of the electricity market price per unit time of the buying and selling. The information processing device according to claim 1 .

9. An information processing method executed by an information processing device, calculating a first estimated value of the first time according to third actual values ​​of a plurality of third time points within a second predetermined period including a period prior to the second time point within a first predetermined period going back from the first time point to be estimated; correcting the first estimated value in accordance with a degree of deviation between the second estimated value at the second time and a second actual value; Information processing methods.

10. On the computer, calculating a first estimated value of the first time according to third actual values ​​of a plurality of third time points within a second predetermined period including a period prior to a second time point within a first predetermined period going back from the first time point to be estimated; correcting the first estimated value in accordance with a degree of deviation between the second estimated value at the second time and a second actual value; Information processing program.

Images