Prediction system and prediction method

Abstract

To accurately predict a bidding curve without difficult-to-obtain information such as power supply configuration information on all power supply facilities that may affect market prices or information on bidding policies of all market participants.SOLUTION: A prediction system, for each pair of a division section number candidate and an approximation function, approximates a portion of a bidding curve to the approximation function, for each division section obtained by division according to the division section number candidate, calculates an approximation evaluation index for the pair, selects a pair of the approximation function and the division section number candidate with the best approximation evaluation index, constructs a parameter model based on a parameter extraction result according to the pair and a factor value for each factor, and predicts the bidding curve using a parameter value predicted using the model and the approximation function in the pair.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present invention relates generally to computer techniques for forecasting. [Background technology]

[0002] In the operation of power generation facilities, various forecast values ​​are used, such as demand forecast, forecast of renewable energy power generation amount, and market price forecast. In particular, when market prices rise sharply, supply may be insufficient relative to demand. Therefore, it is important to predict whether or not there will be a price rise based on the forecast results of market prices and to operate power generation facilities by starting, waiting, and stopping them.

[0003] Patent Document 1 discloses a method of dividing actual market price data into spike price information, where a price increase has occurred, and non-spike price information, where a price increase has not occurred, and predicting prices using a machine learning price algorithm that uses both the spike price information and the non-spike price information.

[0004] Patent Document 2 discloses a method of predicting bid volumes based on the power generation volumes of each power source, such as thermal power, hydroelectric power, atomic power, and renewable energy, and calculating a supply and demand curve to generate market price prediction data.

[0005] Patent Document 3 focuses on the fact that the demand curve and supply curve in the spot market are expressed by the cumulative distribution function of the price distribution of the bid volume, and discloses a method of modeling the bid curve using a normal distribution defined by the average price and standard deviation of buying and selling, respectively. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] JP 2016-212825 A [Patent Document 2] JP 2022-145070 A [Patent Document 3] JP 2017-117356 A Summary of the Invention [Problem to be solved by the invention]

[0007] According to Patent Document 1, the actual data is divided into cases where the market price rises and cases where it does not rise, and each is modeled. However, market price rises usually occur infrequently, and the cause of the price rise may change. For this reason, it is not possible to accumulate enough data to build a model when a price rise occurs, and the accuracy of the model may be low.

[0008] According to Patent Document 2, it is possible to model the causes of price hikes, such as an increase in demand and a decrease in supply, as a bidding curve. However, this technology requires power source configuration information including information on the amount of power generated for all types of power sources, and it is generally difficult to obtain such information, so the feasibility of this technology is low.

[0009] According to Patent Document 3, a bidding curve is predicted from the distribution of bidding prices. However, the distribution of bidding prices is not necessarily a normal distribution, and the distribution shape depends on the bidding policy of each market participant, which is the amount of bidding and the bidding price at which each market participant will bid. It is usually difficult to obtain information about the bidding policy of each market participant. For this reason, the feasibility of this technology is low.

[0010] The present invention has been made in consideration of the above points, and aims to predict a bidding curve with high accuracy without using difficult-to-obtain information such as power source configuration information of all power source facilities that may affect market prices or information on the bidding policies of all market participants. [Means for solving the problem]

[0011] For each pair of a candidate number of division sections and an approximation function, the prediction system approximates a curve portion of the observed bid curve in the division section for each of one or more division sections obtained by division according to the candidate number of division sections to the approximation function, calculates an approximation evaluation index, which is an evaluation index for the pair, and selects a pair of the candidate number of division sections and the approximation function with the best calculated approximation evaluation index. The prediction system constructs a model in which the factor values ​​are input and the parameter values ​​are output based on the parameter extraction results according to the selected pair and the factor values ​​for each factor, and obtains parameter values ​​predicted for the prediction target time for each division section according to the number of division section candidates in the selected pair by inputting the factor values ​​for the prediction target time into the model. The prediction system restores a curve portion for each division section according to the number of division section candidates in the selected pair using the acquired parameter values ​​for each parameter and the approximation function for the selected pair, and predicts the bid curve at the prediction target time based on the curve portion restored for each division section. Effect of the Invention

[0012] According to the present invention, it is possible to accurately predict a bidding curve without difficult-to-obtain information such as power supply configuration information of all power supply facilities that may affect the market price or information on the bidding policies of all market participants. [Brief description of the drawings]

[0013] [Figure 1] FIG. 1 is a diagram showing a configuration of an entire system including a prediction system according to a first embodiment. [Diagram 2] FIG. 1 is a diagram showing an internal configuration of an apparatus according to a first embodiment; [Diagram 3] FIG. 2 is a diagram illustrating a data flow of the prediction system according to the first embodiment. [Figure 4] FIG. 2 is a diagram illustrating a process flow of the prediction system according to the first embodiment. [Diagram 5] FIG. 13 is a diagram showing a data flow of a parameter extraction unit. [Figure 6] FIG. 13 is a diagram showing a processing flow of a parameter extraction unit. [Figure 7] FIG. 4 is a diagram illustrating an overview of processing by a parameter extraction unit. [Figure 8] FIG. 13 is a diagram showing a data flow of a parameter prediction unit. [Figure 9] FIG. 13 is a diagram illustrating an overview of the process of a parameter prediction unit. [Figure 10] FIG. 13 is a diagram showing a data flow of a prediction target value restoration unit. [Figure 11] FIG. 13 is a diagram illustrating an overview of the process of a prediction target value restoration unit. [Figure 12] FIG. 13 is a diagram showing a data flow of a prediction target value verification unit. [Figure 13] FIG. 13 is a diagram illustrating an overview of a process performed by a prediction target value verification unit. [Figure 14A] FIG. 13 is a diagram illustrating an example of an observed bid curve. [Figure 14B] FIG. 13 illustrates an example of an approximated observed bid curve. [Figure 15] FIG. 11 is a diagram illustrating a data flow of a parameter prediction unit according to the second embodiment. [Figure 16] FIG. 13 is a diagram illustrating a data flow of a prediction system according to a third embodiment. [Figure 17] FIG. 13 is a diagram illustrating a data flow of a verification reference value prediction unit according to the third embodiment. [Figure 18] FIG. 13 is a diagram illustrating an overview of a process performed by a verification reference value prediction unit according to the third embodiment. [Figure 19] FIG. 13 is a diagram illustrating a data flow of a prediction system according to a fifth embodiment. [Figure 20] FIG. 13 is a diagram illustrating a data flow of a prediction target value correction unit according to the fifth embodiment. [Figure 21] FIG. 13 is a diagram illustrating a data flow of a prediction system according to a sixth embodiment. [Figure 22] FIG. 13 is a diagram illustrating a data flow of a prediction system according to a seventh embodiment. [Diagram 23] FIG. 23 is a diagram illustrating an overview of a process performed by a bid position estimation unit according to the seventh embodiment. [Figure 24]FIG. 23 is a diagram illustrating a data flow of the prediction system according to the eighth embodiment. [Diagram 25] FIG. 23 is a diagram illustrating a data flow of a correction simulation unit according to the eighth embodiment. [Figure 26] FIG. 23 is a diagram illustrating an overview of a process of a correction simulation unit according to the eighth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] In the following description, an "interface unit" may be one or more interface devices. The one or more interface devices may be at least one of the following: One or more I / O (Input / Output) interface devices. The I / O (Input / Output) interface devices are interface devices to at least one of the I / O devices and a remote display computer. The I / O interface device to the display computer may be a communications interface device. The at least one I / O device may be a user interface device, for example, either an input device such as a keyboard and a pointing device, or an output device such as a display device. One or more communication interface devices. The one or more communication interface devices may be one or more homogeneous communication interface devices (e.g., one or more NICs (Network Interface Cards)) or two or more heterogeneous communication interface devices (e.g., a NIC and an HBA (Host Bus Adapter)).

[0015] In the following description, a "memory" refers to one or more memory devices, typically a primary storage device. At least one memory device in the memory may be a volatile memory device or a non-volatile memory device.

[0016] In the following description, a "persistent storage device" refers to one or more persistent storage devices. A persistent storage device is typically a non-volatile storage device (e.g., an auxiliary storage device), and specifically, for example, a hard disk drive (HDD) or a solid state drive (SSD).

[0017] Also, in the following description, "storage device" may be at least memory, including memory and persistent storage device.

[0018] In the following description, a "processor" refers to one or more processor devices. The at least one processor device is typically a microprocessor device such as a CPU (Central Processing Unit), but may be other types of processor devices such as a GPU (Graphics Processing Unit). The at least one processor device may be a single-core or multi-core. The at least one processor device may be a processor core. The at least one processor device may be a processor device in a broader sense, such as a hardware circuit (e.g., an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit)) that performs part or all of the processing.

[0019] In the following description, functions may be described using the expression "yyy unit", but the functions may be realized by one or more computer programs being executed by a processor, or by one or more hardware circuits (e.g., FPGA or ASIC), or by a combination thereof. When a function is realized by a program being executed by a processor, the function may be at least a part of the processor, since the specified processing is performed using a storage device and / or an interface device, etc., as appropriate. Processing described with a function as the subject may be processing performed by a processor or a device having the processor. A program may be installed from a program source. The program source may be, for example, a program distribution computer or a computer-readable recording medium (e.g., a non-transitory recording medium). The description of each function is an example, and multiple functions may be combined into one function, or one function may be divided into multiple functions.

[0020] An embodiment of the present invention will now be described in detail with reference to the drawings. (1) First embodiment (1-1) Configuration of Prediction System According to the Present Embodiment

[0021] FIG. 1 shows the overall configuration of a system including a prediction system according to this embodiment.

[0022] The data processing system 1 is composed of a prediction system 3 of a prediction system user 2, an observation data storage device 5 of an observation data provider 4, a factor data storage device 7 of a factor data provider 6, and a power generation facility operation system 9 of a power generation facility operator 8. These systems and devices communicate with each other via a communication path 10. The communication path 10 may be a communication network such as a LAN (Local Area Network) or a WAN (Wide Area Network), and is a communication path that connects various devices and terminals that constitute the data processing system 1 so that they can communicate with each other. The power generation facility operation system 9 is a physical computer system (one or more computers having an interface device, a storage device, and a processor connected to them) or a logical computer system based on a physical computer system (for example, a system as a cloud computing service), and uses the prediction results calculated by the prediction system 3 to perform operations including planning and / or control of start-up, standby, or shutdown of power generation facilities such as thermal power generation facilities.

[0023] The observation data storage device 5 stores past observation data related to the prediction target. The observation data is data in the form of bidding curves (buying bid curves and / or selling bid curves) in the electricity trading market. The observation data may be data for all dates and times in the past period, or may be data limited to a specific time, month, season, or other period. The observation data storage device 5 searches for and / or transmits the observation data in response to a data acquisition request from another device.

[0024] The factor data storage device 7 stores factor data (past actual values ​​and / or future predicted values ​​as factor values ​​at each time point) that are input values ​​for calculating the predicted value of the target date and time of the bidding curve (prediction of the bidding curve). The factor data includes, for example, calendar data such as flag values ​​indicating classification such as date, day of the week, weekday, holiday, etc., data on the results of transactions up to the previous day in trading markets such as the spot market and the time-ahead market, predicted data on the demand amount and the amount of power generation by renewable energy for each unit area in which market transactions are conducted, type data including the power generation type such as coal, gas, oil, etc. and the approved output for each power generation facility, start-stop data including the maintenance, inspection, and stoppage and output reduction due to failure of the power generation facility, data on the usage plan of the interconnection lines connecting the power supply areas, and monthly statistics and futures price data of fuel prices. Data on predictions and plans includes data on predictions and plans announced in the past. Furthermore, the data may include past data of the prediction target itself and data converted by, for example, function approximation of the data. The actual number of the factor data storage device 7 is not limited to one, but may be multiple, with different types or periods of factor data stored in each device. The factor data storage device 7 searches for and / or transmits external data in response to a data acquisition request from another device.

[0025] The prediction system 3 uses the observation data and factor data acquired from the observation data storage device 5 and the factor data storage device 7 to calculate the value of an arbitrary target date and time of the prediction target (predicts the bidding curve). (1-2) Device internal configuration

[0026] FIG. 2 shows the configuration of the prediction system 3, the observation data storage device 5, and the factor data storage device 7.

[0027] The prediction system 3 is composed of an input device 32, an output device 33, a communication device 34, a storage device 35, and a CPU (Central Processing Unit) 31 connected to these devices 32 to 35. The prediction system 3 is an information processing system such as a personal computer, a server computer, or a handheld computer. The prediction system 3 may be such a physical computer system (one or more physical computers), or may be a logical computer system based on a physical computer system (for example, a system as a cloud computing service). The input device 32 and the output device 33 may be omitted. The communication device 34 is an example of an interface device, and the CPU 31 is an example of a processor.

[0028] The input device 32 is, for example, a keyboard or a mouse, and the output device 33 is, for example, a display or a printer. The input device 32 and the output device 33 may be included in a remote client with the prediction system 3 as a server.

[0029] The communication device 34 is configured to include a NIC (Network Interface Card) for connecting to, for example, a wireless LAN or a wired LAN. The storage device 35 is a storage medium such as a RAM (Random Access Memory) or a ROM (Read Only Memory). Output results and intermediate results of each processing unit may be output via the output device 33 as appropriate.

[0030] The storage device 35 stores one or more computer programs for implementing functions such as a parameter extraction unit 351, a parameter prediction unit 352, a prediction target value restoration unit 353, and a prediction target value verification unit 354. These functions are implemented by executing these computer programs on the CPU 31. The storage device 35 also has a memory area 355 for storing a prediction value 355A.

[0031] The observation data storage device 5 is composed of at least a communication device 51, a storage device 52, and a CPU 53 connected thereto. The storage device 52 has a storage area 521 for storing data such as observation data. The storage area 521 holds observation data 521A.

[0032] The factor data storage device 7 is composed of at least a communication device 71, a storage device 72, and a CPU 73 connected to them. The storage device 72 has a storage area 721 for storing data such as factor data. The storage area 721 holds factor data 721A. (1-3) Processing and data flow of prediction system 3

[0033] The data flow and processing flow of the prediction system 3 in this embodiment will be described with reference to FIGS.

[0034] First, the data flow of the prediction system 3 in this embodiment will be described with reference to FIG.

[0035] The prediction system 3 obtains the observation data 521A and the factor data 721A from the observation data storage device 5 and the factor data storage device 7, respectively.

[0036] The observation data 521A is input to the parameter extraction unit 531, and the factor data 721A is input to the parameter prediction unit 352 and the prediction target value verification unit 354.

[0037] The parameter extraction unit 351 approximates the input observation data 521A to extract, as parameters, only the features to be modeled from the features of the observation data 521A. The parameters extracted by the parameter extraction unit 351 and the acquired factor data 721A are input to the parameter prediction unit 352.

[0038] The parameter prediction unit 352 models the input parameters (parameters approximating the observation data 521A) and inputs the value of the factor data 721A at the prediction target date and time to the model obtained by modeling. In this way, the parameter prediction unit 352 calculates the value of the parameter at the prediction target date and time. Note that multiple dates and times may be selected as the prediction target date and time. The parameter value at the prediction target date and time calculated by the parameter prediction unit 352 is input to the prediction target value restoration unit 353.

[0039] The prediction target value restoration unit 353 calculates the value of the bidding curve at the prediction target date and time (i.e., predicts the bidding curve) using the parameter values ​​at the prediction target date and time. The bidding curve value calculated by the prediction target value restoration unit 353 is input to the prediction target value verification unit 354 together with the factor data 721A. Note that in this specification, the "bid curve value" and the "bid curve" may be substantially synonymous.

[0040] The forecast target value verification unit 354 calculates a verification reference value by combining the value of the bidding curve calculated by the forecast target value restoration unit 353 and the data of the power generation facility acquired from the factor data 721A, and determines whether the deviation between the verification reference value and the bidding curve calculated by the forecast target value restoration unit 353 is equal to or smaller than a threshold value. Finally, the determined value of the bidding curve is output as a forecast value 355A.

[0041] The predicted market price value 355A is input to the power generation equipment operation system 9, and the result of calculating the market price from the intersection of the bidding curve is used to predict the occurrence of a price surge and is used for operation, including planning and control, of power generation equipment such as thermal power generation equipment and storage equipment such as storage batteries.

[0042] Next, the processing flow of the prediction system 3 will be described with reference to FIG.

[0043] This process is initiated when an input operation is received from a prediction system user 2 who uses the system through an input device 32 provided in the prediction system 3, or when an execution timing separately set in the storage device 35 is reached, and the processes from S301 to S304 are executed by the prediction system 3. Note that the actual process is performed by the CPU 31 provided in the prediction system 3 executing a computer program stored in the storage device 35.

[0044] First, the prediction system 3 acquires the observation data 521A, and calculates parameter values ​​that approximate the observation data 521A (S301).

[0045] Next, the prediction system 3 acquires the approximated parameter values ​​and the factor data 721A, models the parameter values ​​using the factor data 721A (constructs a prediction model), and calculates the parameter values ​​at the prediction target date and time using the values ​​of the factor data 721A at the prediction target date and time (S302).

[0046] Next, the prediction system 3 restores the value of the bid curve at the prediction target date and time based on the parameter values ​​at the prediction target date and time (S303).

[0047] The prediction system 3 calculates a verification reference value from the restored bid curve value at the prediction target date and time and the power generation equipment data included in the factor data 721A, and determines whether the deviation between the verification reference value and the bid curve value is within a threshold value (S304).

[0048] With the above, the processing according to this embodiment is completed.

[0049] Hereinafter, a detailed embodiment of each part will be described. (1-4) Details of each component (1-4-1) Parameter extraction unit 351

[0050] An embodiment of the parameter extraction unit 351 will be described with reference to FIGS.

[0051] FIG. 5 shows the data flow inside the parameter extraction unit 351.

[0052] The parameter extraction unit 351 is made up of an interval approximation unit 3511 and a result selection unit 3512 .

[0053] The interval approximation unit 3511 determines one or more candidates for the number of divided intervals for the observed input curve (input curve identified from the input observation data 521A). For each candidate for the number of divided intervals, the interval approximation unit 3511 divides the observed bid curve into intervals (divided intervals) as many as the candidate for the number of divided intervals, and approximates the curve portion (the curve portion in the divided interval of the observed input curve) for each divided interval. In this way, the interval approximation unit 3511 extracts only the features to be modeled from the features of the input observation data 521A as parameters. The approximation method may be a method of fitting to any function expressed by one or more parameters, such as a linear function such as a polynomial function, or a nonlinear function such as a sigmoid function, a hyperbolic function, or a trigonometric function. The candidate for the number of divided intervals is the number of divided intervals of the observed bid curve. The dimension of the division may be selected arbitrarily. Here, the dimension is an element required to express one point on the bid curve, and in the case of a bid curve, the dimension refers to two elements, the bid amount and the bid price. That is, selecting the dimension of division means selecting whether the bidding curve should be divided by setting a threshold value for the bid amount or the bid price. When dividing into two or more division intervals (that is, when the numerical value specified by the division interval number candidate is an integer of two or more), the division intervals may be equal (the division interval size of each of the two or more division intervals may be the same). Also, the observed bidding curve may be divided at a predetermined division point. Also, the curve portion may be approximated by a different method for each divided division interval. When performing approximation for each divided interval, the interval approximation unit 3511 may interpolate the data divided into the divided intervals. When multiple types of series values ​​are to be divided, such as a buying bid curve and a selling bid curve, different division intervals and approximation methods may be applied to the divided intervals. The interval approximation unit 3511 outputs the parameter extraction results by each division interval number candidate and each approximation method to the result selection unit 3512.

[0054] The result selection unit 3512 evaluates each of the parameter extraction results by the division section number candidates and each approximation method calculated by the section approximation unit 3511, and selects the parameter extraction result to be used for modeling. The parameter extraction result is evaluated by an approximation evaluation index. The approximation evaluation index is calculated as a weighted sum of the error between the observed bidding curve and its approximation result, and the complexity, which is the total value of the number of parameters obtained by the approximation method for each division section. The error may be any type of error, such as an average error or an absolute average error. The weighting values ​​may be any combination including a combination of the same values. The parameter extraction result for which the value of the approximation evaluation index is the smallest is output to the parameter prediction unit 352 as data for modeling.

[0055] The processing flow of the parameter extraction unit 351 will be described with reference to FIG.

[0056] First, the interval approximation unit 3511 selects the number of divided intervals and an approximation method for each divided interval from candidates for the number of divided intervals and candidates for the approximation method (S35101).

[0057] Next, the interval approximation unit 3511 divides the observed bid curve by the selected number of division intervals, and approximates the curve portion for each division interval obtained by the division (the values ​​belonging to that curve portion among the observed data 521A) using the approximation method for that division interval, and extracts parameters (S35102).

[0058] If parameter extraction has been completed for all candidates for the number of divided sections and the candidate approximation methods (S35103: Yes), the process proceeds to S35104; if not completed (S35103: No), the process returns to S35101.

[0059] Next, the result selection unit 3512 selects the parameter extraction result that minimizes the approximate evaluation index (S35104).

[0060] The parameter extraction results calculated by the above process are output to the parameter prediction unit 352, and the process of the parameter extraction unit 351 ends.

[0061] The processing contents of the parameter extraction unit 351 will be described more specifically with reference to FIG.

[0062] As an example, there are a buying bid curve 521A1 and a selling bid curve 521A2, which are expressed as the relationship between the bid amount and the bid price of electricity. The parameter extraction unit 351 performs parameter extraction by approximating each of the buying bid curve 521A1 and the selling bid curve 521A2. In the graph shown in Fig. 7, the dashed line is the observed bid curve, and the solid line is the bid curve as a result of approximating the observed bid curve.

[0063] Candidates for the number of divided sections are "1", "2", and "3". The bidding curve is divided at equal intervals in the direction of the bid price. The only candidate for the approximation method is fitting to a sigmoid function in each divided section. The curve position and curvature of the sigmoid function are extracted as parameters. Note that in this example, the bidding curve is divided by a threshold value of the bid price, but instead of or in addition to the bidding curve, the bidding curve may be divided by a threshold value of the bid amount.

[0064] As a result, parameter extraction results and error index 351A1 for candidate number of division sections "1", parameter extraction results and error index 351A2 for candidate number of division sections "2", and parameter extraction results and error index 351A3 for candidate number of division sections "3" are obtained. From the above results, approximate evaluation indexes of each parameter extraction result when the error and complexity weights are each set to 1 for each candidate number of division sections are calculated. Specifically, it is as follows. According to the code 351A1, the average error is “100” and the complexity is “2” (= the product of the number of divided sections “1” and the number of parameter items “2”), so the approximation evaluation index is 1×100+1×2=102. According to code 351A2, the average error is “20” and the complexity is “4” (= the product of the number of divided sections “2” and the number of parameter items “2”), so the approximation evaluation index is 1×20+1×4=24. According to code 351A3, the average error is “19” and the complexity is “6” (= the product of the number of divided sections “3” and the number of parameter items “2”), so the approximation evaluation index is 1×19+1×6=25.

[0065] Therefore, the approximate evaluation index is smallest when the divided section number candidate is “2”, and therefore the parameter extraction result indicated by reference numeral 351A2 is output to the parameter prediction unit 352. (1-4-2) Parameter prediction unit 352

[0066] An embodiment of the parameter prediction unit 352 will be described with reference to FIGS.

[0067] FIG. 8 shows the data flow of the parameter prediction unit 352.

[0068] The parameter prediction unit 352 is composed of a parameter modeling unit 3521 and a parameter calculation unit 3522 .

[0069] The parameter modeling unit 3521 receives the parameter extraction result calculated by the parameter extraction unit 351 and the factor data 721A from the factor data storage device 7, and models the parameters in the parameter extraction result. The prediction model constructed by modeling may be any type of model, such as a linear model using a multiple regression model or a generalized additive model, a nonlinear model using a polynomial expression, Gaussian process regression, or a neural network, a classification model using a decision tree, a support vector machine, a logistic regression, or clustering, a model using statistics such as the average, median, or mode of values ​​in the observation data 521A, a moving average, an AR (Auto Regressive) model, an ARIMA (Auto Regressive Integrated Moving Average) model, or an analysis model of sequence values, or a combination thereof. The prediction model with the constructed parameters is output to the parameter calculation unit 3522.

[0070] The parameter calculation unit 3522 calculates the parameter value at the prediction target date and time using the model constructed by the parameter modeling unit 3521. There may be multiple prediction target dates and times. The parameter value at the prediction target date and time calculated as above is output to the prediction target value restoration unit 353.

[0071] The processing contents of the parameter prediction unit 352 will be described more specifically with reference to FIG.

[0072] As an example, the buying bid curve and selling bid curve, which are expressed as the relationship between the bid amount and the bid price of electricity, are divided into two intervals using the average of the maximum and minimum bid prices as a threshold, and a prediction model is constructed using the results of extracting parameters for each interval. In the explanation of Fig. 9, it is assumed that the parameters are predicted using this prediction model. The prediction target dates and times are 48 dates and times divided into 30-minute units per day. In addition, in Fig. 9, factor 1, factor 2, ..., factor n are the demand amount, the connection line, ..., and the fuel price.

[0073] First, the parameter extraction result 351A, which is the result of extracting parameters of the past bid curve for each divided section for each date and time, and the factor data 721A1 including one or more factor values ​​associated with each date and time represented by the parameter extraction result 351A, are input to the parameter modeling unit 3521. The parameter modeling unit 3521 uses the parameter extraction result 351A and the factor data 721A1 to construct a prediction model (parameter model) 3521A. The prediction model 3521A is a model that inputs one or more factor values ​​for each date and time and outputs a parameter extraction result predicted for the date and time. A specific example of the prediction model 3521A is as described above. The factor value is a value for a factor. Note that, although FIG. 9 shows an example in which prediction values ​​for 48 time points are calculated from one model, 48 models may be constructed by dividing the model for the 48 time points.

[0074] Next, the parameter calculation unit 3522 inputs the factor data 721A2 including one or more factor values ​​for each prediction target date and time to the prediction model 3521A and calculates the parameter prediction result 352A. The parameter prediction result 352A includes the predicted parameter extraction result (parameter) for each prediction target date and time. The type of factor corresponding to the factor value in the factor data 721A2 is the same as that of the factor data 721A1 used to construct the prediction model 3521A. The parameter prediction result 352A calculated by the above method is output to the prediction target value restoration unit 353. (1-4-3) Prediction target value restoration unit 353

[0075] An embodiment of the prediction target value reconstruction unit 353 will be described with reference to FIG. 10 and FIG.

[0076] FIG. 10 shows a data flow of the prediction target value reconstruction unit 353.

[0077] The prediction target value restoration unit 353 is made up of a division interval restoration unit 3531 and a division interval combination unit 3532 .

[0078] The divided interval restoration unit 3531 receives as input the parameter prediction results (parameter values ​​at each prediction target date and time for each divided interval) calculated by the parameter prediction unit 352, and calculates the value of the bid curve from the parameter value for each divided interval. The calculation of the bid curve value is performed based on the method used when the parameter extraction unit 351 extracted the parameter for that divided interval. In other words, when the parameter extraction unit 351 extracts the parameter by performing function approximation, the divided interval restoration unit 3531 calculates the value of the bid curve by applying the parameter value of that divided interval to the function. The above process is performed for each divided interval. The bid curve value restored for each divided interval is output to the divided interval combination unit 3532.

[0079] The divided interval combining unit 3532 combines the bidding curve values ​​for each divided interval calculated by the divided interval restoration unit 3531, and calculates the bidding curve value for each prediction target date and time. When the divided interval combining unit 3532 calculates the bidding curve for each divided interval by using the divided values ​​as an interpolated value, it extracts only the values ​​within the range of the divided interval from the interpolated values, and combines them with the values ​​of the other divided intervals. The calculated bidding curve value is output to the prediction target value verification unit 354.

[0080] The process contents of the prediction target value reconstruction unit 353 will be described more specifically with reference to FIG.

[0081] As an example, buy and sell bid curves, which are expressed as the relationship between the bid volume and bid price of electricity, are adopted. It is assumed that these bid curves are divided into two sections using the average of the maximum and minimum bid prices as a threshold, and the values ​​of each section are interpolated to the values ​​of the entire section to approximate them using a sigmoid function. The interpolation is performed by duplicating the values ​​of the curve extracted in the divided section, inverting the signs of the bid price and bid volume, and then translating the duplicated curve so that the endpoints of the original curve and the duplicated curve match, and further adding data points consisting of the minimum or maximum bid price and bid volume to match the minimum and maximum bid volume, or deleting data points below the minimum or above the maximum bid volume.

[0082] First, the prediction target value restoration unit 353 receives the predicted value 352A1 including the parameters for each division interval of the buy bid curve, "Buy division interval 1" and "Buy division interval 2", and the parameters for each division interval of the sell bid curve, "Sell division interval 1" and "Sell division interval 2", from the parameter prediction unit 352. The division interval restoration unit 3531 calculates the value of the bid curve for each bid curve and each division interval, using the predicted value 352A1 and the sigmoid function used for approximation, as indicated by the reference numeral 3531A. Here, each calculation result is a calculation result using the value for each division interval as a value obtained by interpolating the value for all division intervals. Next, the division interval combination unit 3532 combines the calculation results of each bid curve and each division interval. Specifically, the division interval combination unit 3532 extracts values ​​that fall within the range of the corresponding division interval from the bid curve calculated as the interpolated value, and combines the values ​​extracted between the buy bid curves and between the sell bid curves. Through the above processing, restoration of the values ​​of the buying bid curve and the selling bid curve at the prediction target date and time is completed, and the values ​​of each bid curve are output to the prediction target value verification unit 354. (1-4-4) Prediction target value verification section

[0083] An embodiment of the prediction target value verification unit 354 will be described with reference to FIGS.

[0084] FIG. 12 shows a data flow of the prediction target value verification unit 354.

[0085] The prediction target value verification unit 354 is composed of a verification reference value calculation unit 3541 and a verification result calculation unit 3542 .

[0086] The verification reference value calculation unit 3541 receives the value of the bidding curve calculated by the prediction target value restoration unit 353 and the factor data 721A as inputs, and calculates a verification reference value for judging the value of the bidding curve. The verification reference value is calculated using estimated values ​​included in the factor data 721A (power generation type such as coal, gas, or oil for each power generation facility, approved output, and estimated power generation unit price for each power generation type) and data on the start / stop status (start / stop status for each power generation facility) at the prediction target date and time. Using data on power generation facilities that are determined to be operable with reference to the start / stop status, the verification reference value calculation unit 3541 calculates a bidding curve that connects the relationship between the power generation amount and the bidding price so that the deviation of the bidding curve calculated by the prediction target value restoration unit 353 is minimized, assuming that the entire amount of power generated is put up for sale bidding, and the value of the bidding curve is set as the verification reference value. The verification reference value is calculated by regarding the authorized output of each power generation facility as the bid volume, the power generation cost as the bid price, and constructing a bidding curve by connecting data points determined by the bid price and the total bid volume. The deviation value may be a known error index such as mean absolute error or mean square error. Note that, in the case where the prediction system user 2 owns the power generation facility to be used in the selling bid, the method is not limited to the above, and the prediction system user 2 may use information related to the bidding method to change the power generation cost and start / stop status data of the power generation facility. The verification reference value calculated as above is output to the verification result calculation unit 3542.

[0087] The verification result calculation unit 3542 receives as input the value of the bidding curve at the prediction target date and time calculated by the prediction target value restoration unit 353 and the verification reference value calculated by the verification reference value calculation unit 3541, and calculates the deviation between the value of the bidding curve at the prediction target date and time and the verification reference value. For the deviation value, an index similar to the error index used by the verification reference value calculation unit 3541 is used. The value of the bidding curve is output as the prediction value 355A.

[0088] The process contents of the prediction target value verification unit 354 will be described more specifically with reference to FIG.

[0089] As an example, it is assumed that, of the buying bid curve and the selling bid curve calculated in FIG. 10, the verification reference value of the selling bid curve is calculated, and the verification result of the selling bid curve is calculated, and then the market price is calculated from the intersection with the buying bid curve. It is also assumed that the power generation facility management data 721B is stored in the storage device 35 (for example, the data 721B may be acquired from the storage area 721 of the factor data storage device 7 and stored in the storage device 35). The power generation facility management data 721B includes at least one of the power generation facility name (an example of the power generation facility ID), the power generation type, the authorized output, and the power generation unit price for each power generation facility (for example, for each power generation facility in the area, including the power generation facility that is the operation target of the power generation facility operation system 9). Note that an example of the "area" referred to here may be an area in which a business operator (for example, a power transmission and distribution business operator) has jurisdiction over power generation facilities (for example, power transmission and distribution facilities).

[0090] First, as shown by the reference numeral 3541A, the prediction target value restoration unit 353 calculates a verification reference value 3541A1 of the selling bid curve by combining power generation facilities so as to reduce the deviation from the calculated selling bid curve 353A. The calculation of the verification reference value 3541A1 is performed by considering the authorized output of each power generation facility as the bid amount and the power generation unit price as the bid price, and constructing a bidding curve by connecting data points determined by the bid price and the total bid amount. In the reference numeral 3541A, the change in the bid price and the addition of the bid amount by the power generation of the power generation facilities AA, BB, CC, and DD is shown. This change is based on the power generation facility management data 721B. As shown by the reference numeral 3542A, the verification result calculation unit 3542 calculates the deviation between the selling bid curve 353A and the verification reference value 3541A1, and judges whether the deviation is within a threshold value. In the example shown in the figure, the deviation "4.7" is calculated as the absolute mean error of the bid price, and the deviation "4.7" is less than the threshold "5.0". Finally, a predicted value 355A including the value of the buying bid curve and the value of the determined selling bid curve is output.

[0091] With the above processing, the calculation processing of the prediction system 3 is completed. (1-5) Description of the Effects of the Present Invention

[0092] The effect of this embodiment will be described with reference to FIG.

[0093] FIG 14A shows an example of observed bid curves (buy bid curves and sell bid curves identified from observed data 521A). FIG 14B shows an example of an approximated observed bid curve. From these figures, it is possible to compare the difficulty of modeling when it is assumed that observed data 521A is directly modeled and when it is assumed that observed data 521A is modeled after approximation.

[0094] 14A, the buying bid curve and the selling bid curve as the observed input curve have curve characteristics such as the bid volume when the bid price falls or rises, the degree of change in the bid price relative to the change in the bid volume after the fall or rise, and a step-like shape in the curve that occurs when the bid price becomes unresponsive to the change in the bid volume midway through the fall or rise process. The step-like shape is, for example, as shown by dotted lines 1301 and 1302.

[0095] The curve characteristics change depending on the bidding method, for example, how much bid volume and at what bid price each market participant places. When directly modeling the observed data 521A, if there is no factor data capable of expressing the curve characteristics, the curve characteristics behave like noise in the observed data 521A, deteriorating the accuracy of modeling and the calculation of the predicted value using the model.

[0096] 14B, according to the bid curve obtained by approximating the observation data 521A, curve features such as a stepped shape are eliminated as shown in dotted frames 1303 and 1304. Therefore, it becomes possible to predict the bid curve with high accuracy even when factors capable of expressing the curve features (factors having a high correlation with the curve features) do not necessarily exist.

[0097] Therefore, by using this embodiment, it is possible to calculate the value of the bidding curve for each of one or more prediction target dates and times of the bidding curve, calculate the market price from the calculation result, and predict whether or not a price surge will occur. Since a price surge and a shortage of supply relative to demand are correlated, it is possible to improve the effect of maintaining the supply and demand balance by using the result of predicting whether or not a price surge will occur in the operation of starting, waiting, and stopping the power generation facility. (2) Second embodiment (addition of data extraction unit 3523 inside parameter prediction unit 352)

[0098] In the first embodiment, the output of the parameter extraction unit 351 and the factor data 721A are input to the parameter modeling unit 3521 of the parameter prediction unit 352. However, this is not limiting, and the data may be divided into one or more data sets by data extraction, and then the data sets may be input to the parameter modeling unit 3521.

[0099] This will be specifically explained using FIG.

[0100] The parameter prediction unit 352 further includes a data extraction unit 3523. The data extraction unit 3523 generates N data sets different from each other by extracting a set of some data from the entire set of data in which the output of the parameter extraction unit 351 and the factor data 721A are associated with each other by a key such as date and time. The parameter modeling unit 3521 constructs (generates) N prediction models by performing modeling for each of the N data sets. The parameter calculation unit 3522 inputs factor data for prediction into the N prediction models to calculate N parameter prediction values. N is an integer of 1 or more. The extraction of the N data sets may be performed by random extraction, or by extracting a part from each cluster obtained by clustering the parameter values, or by extracting a part from each of the factor data classified using one or more types of data. In addition, some data may overlap between the data sets.

[0101] In this embodiment, a value calculated by averaging the parameter values ​​calculated by the parameter calculation unit 3522, calculating a bidding curve after calculating N bidding curves by the prediction target value restoration unit 353, averaging the N bidding curves corrected by the prediction target value correction unit 357, and averaging the predicted value of the market price calculated from the N bidding curves corrected by the prediction target value correction unit 357 may be output as the predicted value 355A. Alternatively, the N bidding curves or the market price or both may be output.

[0102] By calculating N predicted values ​​and taking the average (or taking a value using other statistical methods using N predicted values), error factors such as training errors that arise when modeling parameters can be averaged out among the N models, improving prediction accuracy. In addition, by predicting whether or not there will be a market price rise for each of the N predicted results and calculating the proportion of results in which a rise will occur, it becomes possible to output the probability of a price rise occurring. (3) Third embodiment (addition of verification reference value prediction unit 356)

[0103] In the first embodiment, the prediction target value calculated by the prediction target value restoration unit 353 is used to calculate the verification reference value in the prediction target value verification unit 354, and then it is determined whether the deviation between the prediction target value and the verification reference value is equal to or smaller than a threshold value. However, this is not limited to this, and the verification reference value may be calculated independently and then the deviation may be determined.

[0104] This will be specifically explained using FIG.

[0105] The function of calculating the verification reference value exists as a verification reference value prediction unit 356 independent of the prediction target value verification unit 354. In this embodiment, the prediction target value verification unit 354 has only a verification result calculation unit 3542 in the data flow shown in FIG.

[0106] FIG. 17 shows a data flow of the verification reference value prediction unit 356.

[0107] The verification reference value prediction unit 356 is composed of a verification reference value modeling unit 3561 and a verification reference value calculation unit 3562 .

[0108] The verification reference value modeling unit 3561 receives the observation data 521A and the factor data 721A as inputs, and models the data of the combination of power generation facilities required for the calculation of the verification reference value. The data received as the factor data 721A includes power generation facility data (for example, the power generation facility management data 721B illustrated in FIG. 13 or an extended version thereof) such as the power generation type (such as coal, gas, or oil) and the approved output for each power generation facility, the estimated power generation unit price for each power generation type, and the start / stop status at the prediction target date and time. The verification reference value modeling unit 3561 uses the power generation facility data to create data of the combination of power generation facilities so that the deviation from each bidding curve data included in the observation data 521A is minimized. The format of the data of the combination of power generation facilities may be flag data indicating the start / stop status of each power generation facility, the total power generation amount for each power generation type, or the like. The verification reference value modeling unit 3561 uses the data of the combination of power generation facilities and the factor data to model the combination of power generation facilities. The model constructed by modeling may be any model such as a linear model using a multiple regression model or a generalized additive model, a nonlinear model using a polynomial expression, Gaussian process regression, or a neural network, a classification model using a decision tree, a support vector machine, a logistic regression, or clustering, a model using statistics such as the mean, median, or mode of observed data, an analysis model of series values ​​such as a moving average, an AR (Auto Regressive) model, or an ARIMA (Auto Regressive Integrated Moving Average) model, or a combination of these models. The model constructed by the above method (a model of a combination of power generation facilities) is output to the verification reference value calculation unit 3562.

[0109] The verification reference value calculation unit 3562 calculates the verification reference value at the prediction target date and time using the model constructed by the verification reference value modeling unit 3561. There may be multiple prediction target dates and times. From the results of calculating the combination of power generation facilities, a bidding curve is constructed as the verification reference value. Specifically, the authorized output of each power generation facility is regarded as the bid amount, and the power generation unit price is regarded as the bid price, and the bidding curve is constructed by connecting data points determined by the bid price and the total bid amount. The verification reference value at the prediction target date and time calculated as above is output to the prediction target value verification unit 354.

[0110] A more specific explanation will be given with reference to Fig. 18. Fig. 18 shows an overview of an example of processing for calculating a bidding curve that serves as a verification reference value by predicting a combination of power generation facilities at a prediction target date and time using flag data indicating the start / stop status of each power generation facility as combination data of the power generation facilities (for example, "1" means start / stop, and "0" means stop).

[0111] By providing the verification reference value prediction unit 356 outside the prediction target value verification unit 354 and calculating the correction reference value independently of the prediction target value restoration unit 353, the processing load of modeling and the like increases, but the calculation result of the verification reference value is not affected by the error of the prediction target value restoration unit 353, and it is expected that the accuracy will be improved. In FIG. 18, "AA", "BB", and "CC" are power generation facilities. In the graph 1800, the dashed line is a bidding curve identified from the observation data 521A, and the solid line is a bidding curve reproduced using the power generation facility data. Also, according to the graph 1810, the predicted bidding curve when the power generation facility CC is started at 00:00 and the predicted bidding curve when the power generation facility AA is started 30 minutes later are shown. Further explanation with reference to FIG. 18 will be given later. (4) Fourth embodiment (modification of the process of the parameter extraction unit 351)

[0112] In the first embodiment, the thresholds of the division intervals are set at equal intervals when the parameter extraction unit 351 extracts parameters (the ranges of the division intervals are uniform), but this is not limiting and threshold candidates may be set mechanically. The threshold candidates may be a set of threshold candidates in a number corresponding to the number of division intervals, or may be division threshold candidates obtained by mechanically selecting values ​​that occur frequently in the division direction of the observation data or values ​​that fall within a certain range centered on the value with the high occurrence frequency.

[0113] By mechanically selecting the threshold value of the divided interval, the processing load on the parameter extraction unit 351 increases, but by using the results of more accurate parameter extraction, it becomes possible to more accurately calculate the predicted value. (5) Fifth embodiment (adoption of a prediction target value correction unit 357 instead of the prediction target value verification unit 354)

[0114] In the first embodiment, there is a prediction target value verification unit 354 that determines whether the deviation of the bidding curve value calculated by the prediction target value restoration unit 353 from a verification reference value created using power generation equipment data is equal to or less than a threshold value. However, instead of the prediction target value verification unit 354, a prediction target value correction unit 357 that corrects the value of the bidding curve may be provided.

[0115] 19 shows a data flow of a prediction system 3 equipped with a prediction target value correction unit 357 instead of the prediction target value verification unit 354. The value of the bidding curve calculated by the prediction target value restoration unit 353 is corrected by the prediction target value correction unit 357. The correction result is output as a prediction value 355A.

[0116] FIG. 20 shows a data flow of the prediction target value correction unit 357.

[0117] The prediction target value correction unit 357 is composed of a correction reference value calculation unit 3571 and a correction value calculation unit 3572. The processing content of the correction reference value calculation unit 3571 may be similar to that of the verification reference value calculation unit 3541 constituting the prediction target value verification unit 354, and the processing result is output to the correction value calculation unit 3572 as a correction reference value. The correction value calculation unit 3572 corrects the value of the bidding curve calculated by the prediction target value restoration unit 353 using the correction reference value. The correction method may be any method, and may be, for example, a method of calculating a weighted average of the correction reference value using a weighting coefficient as a combination of any values ​​including the same value and the value of the bidding curve calculated by the prediction target value restoration unit 353. (6) Sixth embodiment (addition of correction reference value prediction unit 358)

[0118] In the sixth embodiment, the correction value is calculated after the correction reference value is calculated independently in the fifth embodiment.

[0119] A specific description will be given with reference to Fig. 21. In Fig. 21, a correction reference value prediction unit 358 that calculates a correction reference value in parallel with the prediction target value restoration unit 353, and a correction value calculation unit 359 that corrects the value of the bidding curve using the correction reference value calculated by the correction reference value prediction unit 358 are provided. The process of the correction reference value prediction unit 358 may be similar to that of the verification reference value prediction unit 356 shown in Fig. 16, and the process result as the correction reference value instead of the verification reference value is output to the correction value calculation unit 359. The process of the correction value calculation unit 359 may be similar to that of the correction value calculation unit 3572 in Fig. 20. (7) Seventh embodiment (addition of bidding position estimation unit 35011)

[0120] In the first embodiment, the result determined by the prediction target value verification unit 354 is output and the process is terminated, but this is not limited to this, and it is also possible to estimate how the prediction system user 2 contributes to the determination of the value of the prediction target.

[0121] This will be specifically explained using FIG.

[0122] The prediction system 3 has a bidding position estimation unit 35011. The bidding position estimation unit 35011 receives as input the factor data 721A and the bidding curve output from the prediction target value verification unit 354. The factor data 721A in this embodiment includes data on the bidding method, such as how much bidding price the prediction system user 2 sets for how much bidding volume. Using the data on the bidding method, the bidding position estimation unit 35011 estimates in which range on the bidding curve output from the prediction target value verification unit 354 the bid of the prediction system user 2 falls. For example, the bidding position estimation unit 35011 specifies, using the bidding curve, how much bidding volume corresponds to the bidding price of the data on the bidding method, and estimates, as the bid of the prediction system user 2, a range of bidding volume with a width corresponding to the bid volume for the bidding method based on the bid volume on the specified bidding curve. The standard may be a statistical value (for example, an average, a maximum value, or a minimum value) of the range of the bid volume. The estimated amount range is visualized by being output via output device 33 and / or added to forecast value 355A.

[0123] Note that the bid position estimation unit 35011 may receive the value of the bid curve from the prediction target value restoration unit 353 instead of or in addition to receiving the value from the prediction target value verification unit 354 .

[0124] By providing the bid position estimation unit 35011 and estimating where on the bidding curve the bid of the prediction system user 2 is located, it becomes possible to perform a simulation to confirm the change in the bidding curve itself and the change in the position on the bidding curve while changing the bidding method. For example, if the prediction system user 2 is a market participant placing a selling bid, the simulation can improve the effect of suppressing a sharp rise in market prices and maintaining the supply and demand balance by resolving supply shortages by increasing the selling bid volume, etc.

[0125] FIG. 23 shows an overview of the process of the bidding position estimation unit 35011.

[0126] As an example, let us assume that prediction system user 2 places a buy bid in the market, and we estimate the bid position of prediction system user 2 on the buy bid curve among the bidding curves calculated in the example of Fig. 11. Consider the case where the bid price is 7.50 and the bid volume is 20. First, the value of the bid volume at the bid price of 7.50 is calculated on the predicted buy bid curve. In this example, the bid volumes on the buy bid curve at the bid price of 7.50 are taken as the average value, and a range of 20 bid volumes is set. On the buy bid curve, the part that falls within the range calculated by the above process will be the bid position of prediction system user 2 in this example. (8) Eighth embodiment (addition of correction simulation unit 35012)

[0127] In the first embodiment, the process is terminated by outputting the result of judgment made by the prediction target value verification unit 354, but this is not limited to the above. A change in the value of the prediction target due to a change in the bid of the prediction system user 2 may be simulated, and the value of the prediction target after the change may be output as the prediction value 355A.

[0128] This will be specifically explained using FIG.

[0129] The prediction system 3 has a correction simulation unit 35012. The correction simulation unit 35012 changes the value of the prediction target by adding or deleting information on the bid amount and bid price by the prediction system user 2, and outputs the changed value as a prediction value 355A. Note that the correction simulation unit 35012 may receive the value of the prediction target from the prediction target value restoration unit 353 instead of or in addition to receiving the value from the prediction target value verification unit 354.

[0130] FIG. 25 shows the data flow of the correction simulation unit 35012.

[0131] The correction simulation unit 35012 is composed of a bidding position estimation unit 350121 and a prediction target value update unit 350122 .

[0132] The bid position estimation unit 350121 receives as input the value of the prediction target output from the prediction target value verification unit 354 (or the prediction target value restoration unit 353) and the factor data 721A. The factor data 721A in this embodiment includes data on the bidding method, such as how much bidding price the prediction system user 2 sets for how much bidding volume. The bid position estimation unit 350121 estimates where on the bidding curve the bid of the prediction system user 2 falls, using information on how much bidding volume and how much bidding price the prediction system user 2 sets and the value of the bidding curve that is the prediction target. The estimation method may be the same as that of the bid position estimation unit 35011 in FIG. 22.

[0133] The prediction target value update unit 350122 receives as input the bid position on the bidding curve of the prediction system user 2 estimated by the bid position estimation unit 350121 and the value of the prediction target output from the prediction target value verification unit 354 (or the prediction target value restoration unit 353). The prediction target value update unit 350122 changes the value of the bidding curve, which is the prediction target, by adding or deleting bidding information determined by a pair of the bid amount and the bid price by the prediction system user 2. In the case of deletion, the prediction target value update unit 350122 deletes from the bidding curve the bidding information corresponding to any of the bid positions on the bidding curve of the prediction system user 2 estimated by the bid position estimation unit 350121. In the case of addition, the prediction target value update unit 350122 adds bidding information and inserts the added amount into the bidding curve. When processing the selling bid curve, the forecast target value update unit 350122 also receives as input the factor data 721A including the power generation type (such as coal, gas, or oil) and the approved output for each power generation facility, the estimated power generation unit price for each power generation type, and the start / stop status data at the forecast target date and time, and assumes that the power generation facility scheduled to be shut down at the forecast target date and time will be operated, and may set the power generation unit price of the power generation facility in question as the bid price and the amount of the approved output to be bid as the bid amount as additional bid information. The forecast target value update unit 350122 repeats the above processing until the market price determined by the buying bid curve and the selling bid curve falls within a certain threshold value.

[0134] The process of the correction simulation unit 35012 will be described more specifically with reference to FIG.

[0135] As an example, assume a case where bid information is added or deleted by targeting the sell bid curve among the values ​​calculated by the prediction target value restoration unit 353. According to the example shown in Fig. 26, the bid of the prediction system user 2 is estimated by the bid position estimation unit 350121 to be at bid positions 1, 2, and 3 on the sell bid curve.

[0136] An example of deleting bid information is as shown by the reference symbol 2501. The correction simulation unit 35012 deletes the portion of the selling bid curve that corresponds to bid position 3, and translates the portion of the curve to the right of bid position 3 to the left by the bid amount of bid position 3.

[0137] An example of adding bid information is as shown by the reference symbol 2502. The correction simulation unit 35012 assumes a case where a power generation facility CC that was scheduled to be shut down is started up and a new selling bid is made with a power generation unit price of 33 yen as the bid price. The correction simulation unit 35012 identifies the position on the original buying bid curve that corresponds to 33 yen, and inserts a constant function having a width equivalent to the amount of power generation of the power generation facility CC at that position.

[0138] As described above, the corrected simulation unit 35012 adds or deletes the bid information of the prediction system user 2 to or from the predicted bid curve, and outputs the processing result as the predicted value 355A.

[0139] Although several embodiments of the present invention have been described above, these are merely examples for the purpose of explaining the present invention, and the scope of the present invention is not limited to these embodiments. The present invention can be embodied in various other forms. For example, a form in which two or more of the above-described embodiments are used in combination may be adopted.

[0140] The above description can be summarized, for example, as follows. The following summary may include supplementary explanations and explanations of variations of the above description. Note that in the above embodiment and the following summary, the bidding curve is a bidding curve (bidding curve related to electricity) that represents the relationship between the bid amount of electricity and the bid price (price per unit amount of electricity), but the present invention may be applied to predicting a bidding curve related to something other than electricity (for example, a bidding curve for an object that can be provided and bought and sold based on a group of facilities including one or more facilities that can be operated by a facility operation system (a computer system that operates facilities)). Furthermore, an object other than a bidding curve may be the prediction target.

[0141] The prediction system 3 includes an interface device (e.g., a communication device 34), a storage device 35, and a processor (e.g., a CPU 31) connected thereto. The interface device is communicatively connected to an observation data storage device 5 that stores observation data, a factor data storage device 7 that stores factor data, and a power supply equipment operation system (e.g., a power generation equipment operation system 9) that performs operations including at least one of planning and controlling the start, standby, or stop of one or more power supply equipment (e.g., one or more power generation equipment). The observation data 521A includes data of a bidding curve representing a relationship between the bid amount of power and the bid price per unit bid amount for each time. The factor data 721A includes a factor value for each of one or more factors that may affect the bidding relationship for each time. There are one or more approximation functions (e.g., functions such as a linear function and a nonlinear function), each of which is one or more functions represented by one or more parameters.

[0142] The processor identifies a bidding curve at each time from the observation data 521A received from the observation data storage device 5. The processor divides the observed bidding curve at each time, which is the bidding curve at each time, into one or more divided sections according to the divided section candidate for each of the one or more divided section number candidates. The divided section number candidate is a policy including how many divided sections the bidding curve is divided into, and may include other policies in addition to the number of divided sections, for example. For each pair of the divided section number candidate and the approximation function for the observed bidding curve at each time, the processor approximates the curve part in the divided section of the observed bidding curve to the approximation function for each pair of the divided section number candidate and the approximation function for the observed bidding curve at each time, which is an evaluation index of the pair. The processor selects the pair of the divided section number candidate and the approximation function for which the calculated approximation evaluation index is the best. The processor constructs a prediction model 3521A (parameter model) which is a model in which the factor value for each factor is input and the parameter value for each parameter is output based on the parameter extraction result according to the selected set and the factor value for each factor in the factor data 721A received from the factor data storage device 7. The parameter extraction result (see, for example, reference numeral 351A in FIG. 9) includes, for each divided section according to the number of divided section candidates in the selected set, a parameter value obtained from an observed bid curve approximated to the approximation function for each parameter of the approximation function in the selected set for each time. The processor inputs the factor value for each factor at the prediction target time (for example, each time in the prediction target period) to the constructed prediction model, thereby obtaining a predicted parameter value for each parameter at the prediction target time for each divided section according to the number of divided section candidates in the selected set. The processor reconstructs a curve portion for each division interval according to the number of division interval candidates in the selected set using the acquired parameter values ​​for each parameter and the approximation function in the selected set (see, for example, reference numeral 3531A in FIG. 11), and predicts a bid curve at the prediction target time based on the curve portion reconstructed for each division interval (see, for example, reference numeral 3532A in FIG. 11).The processor stores the predicted bidding curve data, which is data representing a predicted bidding curve, in the storage device 35. The processor transmits the predicted bidding curve data or data based on the predicted bidding curve data to the power supply equipment operation system. The "data based on the predicted bidding curve data" may be power generation plan data representing a power generation plan, for example, power generation plan data representing a power generation plan of when and how to operate which power supply equipment in the prediction target period based on the power generation amount (e.g., authorized output) and power generation unit price (price of electricity per unit amount of electricity) of each power supply equipment that can be an operation target and a predicted bidding curve for each of one or more prediction target times in the prediction target period. Furthermore, the "data based on the predicted bidding curve data" may be control instruction data, which is a control instruction based on the power generation plan, instead of or in addition to the power generation plan data. The transmission destination equipment and transmission timing of the control instruction data follow the power generation plan.

[0143] In this way, by functionally approximating the observed bidding curve (the bidding curve performance), only a part of the characteristics of the curve shape of the bidding curve is selectively extracted, and by modeling only the extracted characteristics, the bidding curve can be accurately predicted without difficult-to-obtain information such as power supply configuration information of all power supply facilities that may affect the market price and information on the bidding policy of all market participants. Specifically, for example, even if the factor data 721A does not contain factor values ​​of factors that are highly correlated with the bidding curve, the prediction accuracy can be improved. As a result, for example, it is possible to accurately predict the occurrence of a market price surge (for example, to predict a shortage of power supply), and it is expected that the power supply facilities will be operated appropriately and the supply and demand balance will be maintained. For example, the prediction system 3 may be provided for each business operator such as a retail business operator or a power generation business operator (business operator that can participate in the electricity wholesale market). In addition, both the observation data and the factor data may be publicly available data.

[0144] The processor may generate a verification reference curve (e.g., verification reference value 354A1) that is a curve with a minimum deviation from a predicted bidding curve and a bid amount and a bid price based on the power generation amount and the power generation unit price of each of one or more power supply facilities (based on the power generation facility management data 721B, for example), and determine whether or not the deviation between the generated verification reference curve and the predicted bidding curve is equal to or less than a threshold. The bidding curve represented by the predicted bidding curve data may be a bidding curve determined to have a deviation from the generated verification reference curve equal to or less than a threshold. One factor that affects the bidding curve is the power supply and demand balance, which may depend on the operation of each power supply facility. For this reason, a bidding curve that exceeds the operation range of each power supply facility (a bidding curve whose deviation from the verification reference curve exceeds a threshold) is a bidding curve with low feasibility, and it is possible to avoid using such a bidding curve in bidding or facility operation. Note that such a verification reference curve may be a curve for correcting the bidding curve. That is, the processor may correct the predicted bidding curve using the generated verification reference curve (here, the curve for correction). This is expected to improve the prediction accuracy. For example, the corrected bid curve may be a weighted average of the corrected reference curve and the predicted bid curve.

[0145] For each pair of the division section number candidate and the approximation function for the observed bid curve at each time, the calculated approximation evaluation index is based on at least one of a first index which is an error between the observed bid curve and the observed bid curve approximated to the approximation function, and a second index which is an index based on the number of division sections obtained by division according to the division section number candidate and the number of parameters of the approximation function. This is expected to accurately evaluate the appropriateness of the approximation according to the pair of the division section number candidate and the approximation function. For example, the approximation evaluation index is defined as a weighted sum of the approximation error and the complexity, and the coefficients representing the respective weights can be any value including the same value of 0 or more.

[0146] The processor may divide the range from the minimum value to the maximum value in at least one dimension of the bid curve, which is the bid price and the bid volume, according to the number of division interval candidates, thereby dividing the observed bid curve into division intervals whose number corresponds to the number of division interval candidates. Note that when the number of division interval candidates is "1", no division is performed, and as a result, there may be one division interval that includes the entire observed bid curve. Also, for each number of division interval candidates, a threshold value that determines the boundary of the division interval, which is the threshold value for at least one dimension of the bid volume and the bid price, may be based on the number of division intervals according to the number of division interval candidates and the occurrence frequency of the value of the dimension in the observed bid curve at multiple times (the size of each division interval does not need to be uniform).

[0147] The processor may create N data sets (N is an integer equal to or greater than 1) each composed of a part of the parameter extraction result and a part of the factor data 721A from the parameter extraction result and the factor data 721A, and may construct a prediction model using the N data sets with the factor value as an input and the parameter value as an output for each of the N data sets. In one data set, the part of the parameter extraction result and the part of the factor data 721A may have the same time set (one or more times), and the parameter value for each parameter and the factor value for each factor may be included in the data set for each of the same times. Alternatively, a data set may exist for each divided section, and for example, one data set may include a parameter value for each parameter at each time and a factor value for each factor at each time for the divided section corresponding to the data set. One prediction model may be constructed based on one data set. The processor may obtain a parameter value for the prediction target time for each of the N constructed prediction models by inputting a factor value for the prediction target time as a part of the time of the prediction target period (for example, a future period). The bidding curve can be predicted based on the output of N prediction models. This is expected to improve prediction accuracy by smoothing out error factors such as training errors that occur when modeling parameters across N prediction models.

[0148] The processor may create power supply equipment combination data (hereinafter, the power supply equipment combination data 18 illustrated in FIG. 18 is taken as an example) that minimizes the deviation from the observed bidding curve based on power supply equipment management data (see, for example, reference numeral 721B in FIG. 13) including the power generation amount and power generation cost for each power supply equipment. The power supply equipment combination data 18 may be data that represents a power line equipment combination (one or more power supply equipment) for each time represented by the observation data 521A. The processor may construct a verification model 19 (see FIG. 18) that is a model that inputs the factor value for each factor and outputs the power supply equipment combination based on the power supply equipment combination data and the factor value for each factor at each time. The processor may obtain a power supply equipment combination predicted for the prediction target time by inputting the factor value for each factor for the prediction target time into the constructed verification model 19. The processor may predict a verification reference curve (see, for example, reference numeral 1810 in FIG. 18 ), which is a curve with the bid amount and bid price as dimensions, for the prediction target time based on the power generation amount and power generation unit price (for example, the power generation amount and power generation unit price specified from the power generation equipment management data) of each of the one or more power generation equipment in the acquired power generation equipment combination. The processor may determine whether or not the deviation between the predicted verification reference curve and the predicted bidding curve is equal to or less than a threshold value. The bidding curve represented by the predicted bidding curve data may be a bidding curve determined to have a deviation from the predicted verification reference curve equal to or less than a threshold value. This prevents the calculation result of the verification reference curve from being affected by the error of the predicted bidding curve, and is expected to improve the prediction accuracy. Note that the power generation equipment combination data 18 may include data indicating how each power generation equipment is operated at each time (for example, flag data including a flag indicating a status such as start or stop for each power generation equipment), and / or data indicating the total power generation amount of one or more power generation equipment. Note that such a verification model may be a correction model, and such a verification reference curve may be a correction curve for the bidding curve. That is, the processor may correct the predicted bid curve using the generated verification reference curve (here, the correction curve). This is expected to improve the prediction accuracy. For example, the corrected bid curve may be a weighted average of the correction reference curve and the predicted bid curve.

[0149] The processor may specify, from the predicted bidding curve, a bidding volume corresponding to the bidding price specified from the bidding policy data (e.g., data included in the factor data 721A) that represents a bidding policy, which is a combination of a bid volume and a bid price, for a user of the prediction system 3 (prediction system user 2). The processor may estimate, as a position of the user's bid on the predicted bidding curve, a range defined by a first bid volume that is the specified bid volume and a second bid volume that is a bid volume that corresponds to the specified bid price in the bidding policy. The processor may provide the estimated bid position to the user (for example, may display it on a display device that is an example of the output device 33).

[0150] The processor may correct the predicted bidding curve based on removing any of the estimated one or more bid positions (e.g., see reference number 2501 in FIG. 26) or adding data as a combination of new bid quantity and bid price (e.g., see reference number 2502 in FIG. 26).

[0151] One or more parameters of any of the one or more functions (e.g., a sigmoid function) may include a pair of a curve position and a curvature. The curve position may be a position on the bid curve that satisfies any of the one or more conditions (e.g., an inflection point). The curvature may be the curvature at the curve position. [Explanation of symbols]

[0152] 3... Prediction system, 5... Observation data storage device, 7... Factor data storage device, 9... Power generation facility operation system

Claims

1. an interface device communicably connected to an observation data storage device that stores observation data, a factor data storage device that stores factor data, and a power supply equipment operation system that performs operations including at least one of planning and controlling start-up, standby, or shutdown of one or more power supply equipment; A storage device; a processor connected to the interface device and the storage device; Equipped with the observation data includes, for each time, data of a bidding curve that represents a relationship between a bid amount of electricity and a bid price per unit bid amount; the factor data including, for each time, a factor value for each of one or more factors that may affect a bidding relationship; there are one or more approximating functions, each of which is one or more functions described by one or more parameters; The processor determines a bid curve for each time from the observation data received from the observation data storage device; The processor divides an observed bid curve at each time, which is a bid curve at each time specified, into one or more divided intervals according to one or more divided interval number candidates, each of which is a policy including how many divided intervals the bid curve is divided into; the processor, for each pair of a candidate number of division sections and an approximation function for the observed bid curve at each time, approximates a curve portion of the observed bid curve in the one or more division sections obtained by division according to the candidate number of division sections to the approximation function; the processor calculates, for each pair of a candidate number of division sections and an approximation function for the observed bid curve at each time, an approximation evaluation index which is an evaluation index of the pair; the processor selects a combination of a candidate number of divided sections and an approximation function that provides the best calculated approximation evaluation index; The processor constructs a prediction model, which is a model in which the factor value for each factor is input and the parameter value for each parameter is output, based on the parameter extraction result according to the selected set and the factor value for each factor at each time in the factor data received from the factor data storage device; the parameter extraction result includes, for each division interval according to the number of division interval candidates in the selected set, for each time, a parameter value obtained from an observed bid curve approximated to the approximation function for each parameter of the approximation function in the selected set; The processor inputs a factor value for each factor at the prediction target time into the constructed prediction model, thereby obtaining a predicted parameter value for each parameter at the prediction target time for each divided interval according to the number of divided interval candidates in the selected set; The processor reconstructs a curve portion for each divided section according to the number of divided section candidates in the selected set, using the acquired parameter values ​​for each parameter and the approximation function in the selected set, and predicts a bid curve at the prediction target time based on the curve portion reconstructed for each divided section; The processor stores predicted bid curve data in the storage device, the predicted bid curve data being data representing a predicted bid curve; The processor transmits the forecast bid curve data or data based on the forecast bid curve data to the power supply equipment operation system. Prediction system.

2. The processor, Generate a verification reference curve, which is a curve having a minimum deviation from the predicted bidding curve and is a curve having dimensions of a bid amount and a bid price, based on the power generation amount and the power generation unit price of each of the one or more power supply facilities; determining whether the deviation between the generated verification reference curve and the predicted bid curve is less than or equal to a threshold; The bid curve represented by the predicted bid curve data is a bid curve whose deviation from the generated verification reference curve is determined to be equal to or less than the threshold value. The prediction system of claim 1 .

3. For the observed bid curve at each time, the calculated approximation evaluation index for each pair of the candidate number of division sections and the approximation function is based on at least one of a first index which is an error between the observed bid curve and the observed bid curve approximated by the approximation function, and a second index which is an index based on the number of division sections obtained by division according to the candidate number of division sections and the number of parameters of the approximation function; The prediction system of claim 1 .

4. the processor divides a range from a minimum value to a maximum value in at least one dimension of the bid curve, which is a bid price and a bid amount, in accordance with the number of division interval candidates, thereby dividing the observed bid curve into a number of division intervals in accordance with the number of division interval candidates; The prediction system of claim 1 .

5. The processor, creating N data sets each composed of a part of the parameter extraction result and a part of the factor data from the parameter extraction result and the factor data (N is an integer equal to or greater than 1); For each of the N data sets, constructing a predictive model using the data set, the predictive model having factor values ​​as inputs and parameter values ​​as outputs; For each of the N constructed prediction models, a factor value for a prediction target time, which is a part of a prediction target period, is input to obtain a parameter value for the prediction target time. The prediction system of claim 1 .

6. The processor creates power supply equipment combination data that minimizes deviation from the observed bidding curve based on power supply equipment management data including the amount of power generated and the unit price of power generation for each power supply equipment, The power supply equipment combination data is data representing a power line equipment combination for each time represented by the observation data, The processor, constructing a verification model based on the power supply equipment combination data and the factor values ​​for each factor at each time, the verification model being a model in which the factor values ​​for each factor are input and the power supply equipment combination is output; A power supply equipment combination for a target time is obtained by inputting a factor value for each factor for the target time of prediction into the constructed verification model; predicting a verification reference curve, which is a curve having dimensions of a bid amount and a bid price, for the prediction target time based on the power generation amount and the power generation unit price of each of the one or more power supply facilities in the obtained power supply facility combination; determining whether the deviation between the predicted verification reference curve and the predicted bid curve is less than or equal to a threshold; The bid curve represented by the predicted bid curve data is a bid curve whose deviation from the predicted verification reference curve is determined to be equal to or less than the threshold value. The prediction system of claim 1 .

7. The power supply equipment combination data includes data representing which power supply equipment is being operated and how, at each of the times, and / or data representing a total amount of power generation of one or more power generation equipment. The prediction system of claim 6.

8. For each of the number of division interval candidates, a threshold value for determining the boundaries of the division interval, the threshold value for at least one dimension of the bid volume and the bid price, is based on the number of division intervals according to the number of division interval candidates and the occurrence frequency of the value of the dimension in the observed bidding curve at multiple times; The prediction system of claim 4.

9. The processor, Generate a correction reference curve, which is a curve having a minimum deviation from the predicted bidding curve and is a curve having dimensions of a bid amount and a bid price, based on the power generation amount and the power generation unit price of each of the one or more power supply facilities; calibrating the predicted bid curve using the generated calibration reference curve; The prediction system of claim 1 .

10. The processor creates power supply equipment combination data that minimizes deviation from the observed bidding curve based on power supply equipment management data including the amount of power generated and the unit price of power generation for each power supply equipment, The power supply equipment combination data is data representing a power line equipment combination for each time represented by the observation data, The processor, constructing a correction model that is a model in which the factor value for each factor is input and the power supply equipment combination is output based on the power supply equipment combination data and the factor value for each factor at each time; A power supply equipment combination for a target time is obtained by inputting a factor value for each factor for the target time of prediction into the constructed correction model; generating a correction reference curve, which is a curve having dimensions of a bid amount and a bid price for the prediction target time, based on the power generation amount and the power generation unit price of each of the one or more power supply facilities in the acquired power supply facility combination; calibrating the predicted bid curve using the generated calibration reference curve; The prediction system of claim 1 .

11. The processor, Based on bidding policy data representing a bidding policy, which is a combination of a bidding volume and a bidding price, for a user of the prediction system, a bidding volume corresponding to the bidding price identified from the bidding policy data is identified from the predicted bidding curve; A range defined by a first bid volume, which is the specified bid volume, and a second bid volume, which is a bid volume corresponding to the specified bid price in the bidding policy, is estimated as a position of the user's bid on the predicted bidding curve; The prediction system of claim 1 .

12. the processor corrects the predicted bid curve based on removing any of the estimated one or more bid positions or adding new bid volume and bid price combination data. The prediction system of claim 11.

13. one or more parameters of any one of the one or more functions include a pair of a curve position and a curvature; the curve position is a position on the bid curve that satisfies any one of one or more conditions; Curvature is the curvature at the curve position. The prediction system of claim 1 .

14. The computer determines a bid curve at each time from the observation data received from the observation data storage device; the observation data includes, for each time, data of a bidding curve that represents a relationship between a bid amount of electricity and a bid price per unit bid amount; The computer divides an observed bid curve at each time, which is a bid curve at each specified time, into one or more divided intervals according to one or more divided interval number candidates, each of which is a policy including how many divided intervals the bid curve is divided into; the computer, for each pair of a candidate number of division sections and an approximation function for the observed bid curve at each time, approximates a curve portion of the observed bid curve in the one or more division sections obtained by dividing the observed bid curve in accordance with the candidate number of division sections to the approximation function; the computer calculates, for each pair of a candidate number of division sections and an approximation function for the observed bid curve at each time, an approximation evaluation index which is an evaluation index of the pair; there are one or more approximating functions, each of which is one or more functions described by one or more parameters; the computer selects a combination of a candidate number of divided sections and an approximation function that has the best calculated approximation evaluation index; the computer constructs a prediction model, which is a model in which the factor value for each factor is input and the parameter value for each parameter is output, based on the parameter extraction result according to the selected set and the factor value for each factor at each time in the factor data received from the factor data storage device; the factor data including, for each time, a factor value for each of one or more factors that may affect a bidding relationship; the parameter extraction result includes, for each division interval according to the number of division interval candidates in the selected set, for each time, a parameter value obtained from an observed bid curve approximated to the approximation function for each parameter of the approximation function in the selected set; The computer inputs factor values ​​for each factor at the prediction target time into the constructed prediction model, thereby obtaining predicted parameter values ​​for each parameter at the prediction target time for each divided interval according to the number of divided interval candidates in the selected set; The computer restores a curve portion for each divided section according to the number of divided section candidates in the selected set, using the acquired parameter values ​​for each parameter and the approximation function in the selected set, and predicts a bid curve at the prediction target time based on the curve portion restored for each divided section; The computer stores predicted bid curve data, which is data representing a predicted bid curve, in the storage device; The computer transmits the forecast bidding curve data or data based on the forecast bidding curve data to a power supply equipment operation system that performs operations including at least one of planning and controlling start-up, standby, or shutdown of one or more power supply equipment; Forecasting methods.

15. a prediction system communicably connected to an observation data storage device that stores observation data, a factor data storage device that stores factor data, and a power supply equipment operation system that performs operations including at least one of planning and controlling the start-up, standby, or stoppage of one or more power supply equipment; The power supply equipment operation system; Equipped with the observation data includes, for each time, data of a bidding curve that represents a relationship between a bid amount of electricity and a bid price per unit bid amount; the factor data including, for each time, a factor value for each of one or more factors that may affect a bidding relationship; there are one or more approximating functions, each of which is one or more functions described by one or more parameters; The prediction system determines a bidding curve for each time from the observation data received from the observation data storage device; The prediction system divides an observed bid curve at each time, which is a bid curve at each specified time, into one or more divided intervals according to one or more divided interval number candidates, each of which is a policy including the number of divided intervals into which the bid curve is divided; the prediction system, for each pair of a candidate number of division sections and an approximation function for the observed bid curve at each time, approximates a curve portion of the observed bid curve in the one or more division sections obtained by division according to the candidate number of division sections to the approximation function; the prediction system calculates, for each pair of a candidate number of division sections and an approximation function for the observed bid curve at each time, an approximation evaluation index which is an evaluation index of the pair; the prediction system selects a combination of a candidate number of divided sections and an approximation function that has the best calculated approximation evaluation index; the prediction system constructs a prediction model, which is a model in which the factor value for each factor is input and the parameter value for each parameter is output, based on the parameter extraction result according to the selected set and the factor value for each factor at each time in the factor data received from the factor data storage device; the parameter extraction result includes, for each division interval according to the number of division interval candidates in the selected set, for each time, a parameter value obtained from an observed bid curve approximated to the approximation function for each parameter of the approximation function in the selected set; The prediction system inputs a factor value for each factor at a time to be predicted into the constructed prediction model, thereby obtaining a predicted parameter value for each parameter at the time to be predicted for each divided interval according to the number of divided interval candidates in the selected set; the prediction system restores a curve portion for each divided section according to the number of divided section candidates in the selected set, using the acquired parameter values ​​for each parameter and an approximation function in the selected set, and predicts a bid curve at the prediction target time based on the curve portion restored for each divided section; the prediction system stores predicted bid curve data, which is data representing a predicted bid curve, in the storage device; The prediction system transmits the forecast bidding curve data or data based on the forecast bidding curve data to the power supply equipment operation system; The power supply equipment operation system operates the one or more power supply equipment based on the data from the prediction system. Data processing system.

Images