Information processing device, information processing method and program

Abstract

To calculate an appropriate bid amount of a transaction object with high accuracy in markets of electric power or the like.SOLUTION: An information processing device is provided with a processing part. The processing part inputs a transaction amount in a first market, and order book including a bid amount and a bid price of the transaction in a second market different from the first market, calculates a plurality of first contract prices each corresponding to a plurality of first input amounts by using a price estimation model learned so as to output contract prices of the transaction object in the second market, selects a plurality of estimated prices representing first contract prices each corresponding to a plurality of risk levels from a plurality of first contract prices, and calculates a plurality of estimated bid amounts each corresponding to the plurality of estimated prices.SELECTED DRAWING: Figure 1

Description

[Technical field]

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

[0002] There is a known electricity trading market for trading electricity generated by power generation companies and others. For example, in Japan, electricity is traded in multiple markets set up by the Japan Electric Power Exchange (JEPX). The multiple markets include, for example, a spot market and an advance market. The spot market is a market in which electricity to be sold or purchased the next day is bid on by the day before, and the transaction is concluded. The advance market (intraday market) is a market that can be used up until just before actual supply and demand.

[0003] Regarding the forward market, a technology has been proposed to generate a trading plan from the output of a model that simulates future price fluctuations using features in the dependencies between products. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2020-184246 A Summary of the Invention [Problem to be solved by the invention]

[0005] An object of the present invention is to provide an information processing device, an information processing method, and a program that are capable of determining with a high degree of accuracy an appropriate bid amount for a trading object in a market such as electricity. [Means for solving the problem]

[0006] An information processing device according to an embodiment includes a processing unit, which inputs a trading volume of a trading object in a first market and order book information including a bid volume and a bid price of the trading object in a second market different from the first market, calculates a plurality of first contract prices corresponding to the plurality of first bid volumes using a price estimation model trained to output a contract price of the trading object in the second market, selects a plurality of estimated prices representing first contract prices corresponding to a plurality of risk levels, respectively, from the plurality of first contract prices, and calculates a plurality of estimated bid volumes corresponding to the plurality of estimated prices, respectively. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a block diagram of a configuration of an information processing apparatus according to an embodiment. [Diagram 2] FIG. 4 is a diagram showing an example of a data structure of contract data. [Diagram 3] FIG. 13 is a diagram showing an example of a data structure of board information. [Figure 4] FIG. 1 shows an example of a training dataset. [Diagram 5] FIG. 1 is a diagram showing an overview of processing by a price estimation model. [Figure 6] FIG. 4 is a diagram showing an overview of processing performed by a calculation unit. [Figure 7] FIG. 13 is a diagram showing an example of a method of operating a bidding plan calculation process. [Figure 8] 11 is a flowchart of a bidding plan calculation process according to an embodiment. [Figure 9] FIG. 2 is a hardware configuration diagram of the information processing apparatus according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] A preferred embodiment of the information processing device according to the present invention will be described in detail below with reference to the accompanying drawings. In the following, an example will be described in which the market and the trading object in the market are an electricity trading market and electricity, but the market and the trading object are not limited to these. In this example, the trading volume, which represents the amount of the trading object (electricity) traded in the market, corresponds to the amount of electricity sold or purchased.

[0009] As mentioned above, the Japan Electric Power Exchange (JEPX) operates two trading markets: the spot market and the hourly market. In the spot market, trading is conducted based on the day-before estimate (forecast) of power generation. On the other hand, in the hourly market, trading is conducted to minimize the difference (gap) with the demand plan according to the update of the estimated power generation amount, while at the same time maximizing profits. With the spread of renewable energy, it is required to determine an appropriate trading plan (bidding plan) in the hourly market in order to pursue high profits while ensuring a stable supply of electricity.

[0010] The pre-hours market uses the continuous bidding method. In this method, buy and sell bids are lined up in order of price, and a transaction is made when the price conditions are met. This can cause prices to fluctuate suddenly. In order to take into account these factors that cause sudden price fluctuations and not only reduce the imbalance (the difference between the amount of electricity demanded and the amount of electricity supplied) but also ensure high profits, technology is needed to optimally determine the bid volume, bid price, and bid timing.

[0011] As described above, a technology has been proposed that generates a trading plan from the output of a model that simulates future price fluctuations using features in the dependency relationships between products. Specifically, in this technology, price data for products in different markets is input, features in the dependency relationships between products are extracted, and a model that estimates future prices is created. In addition, price fluctuations are simulated by recursively calling this model, and a trading plan is generated that takes into account price fluctuations and the adjustment capacity of power generation sources, placing buy orders at times when the estimated price is low and selling orders at times when the estimated price is high.

[0012] However, the factors that affect the price in the pre-hours market are not limited to the dependency between products, and more diverse factors must be considered. Furthermore, in the above-mentioned technology, the uncertain factors of the winning risk are not taken into account when planning a transaction, so there is a concern that an imbalance will occur and profits will decrease if a bid is not won. In the following, a bid that is not won will be referred to as a failed bid. A bid that is not won is also called a failed bid, a failed bid, or a failed contract.

[0013] In this embodiment, the following functions make it possible to more accurately determine an appropriate bid amount for the trading subject (electricity) in the electricity market. (F1) The contract price is estimated using a price estimation model that uses the time-ahead market information as input. (F2) The risk of non-fulfillment of the bid (risk of non-fulfillment of the bid) is estimated from the information on the order book in the time-ahead market, and the most appropriate bidding plan (bidding volume and bid price) is calculated from among multiple bidding plans in combination with a price estimation model.

[0014] As a result, it is possible to optimize the trading plan (bidding plan) in the ahead-of-hours market, capture price fluctuations in the ahead-of-hours market more accurately, and ensure high profits while reducing imbalances. In other words, it is possible to improve efficiency and achieve stable supply in the electricity trading market.

[0015] In the following embodiment, a case where electricity is sold in the hourly market (the trading volume is the power selling volume) will be mainly described as an example, but a similar procedure can be applied to a case where electricity is purchased (the trading volume is the power purchasing volume). The electricity sold in the hourly market may be obtained in any manner, but it may be, for example, electricity generated by a power generation company or electricity obtained from power sources bundled by a power aggregator.

[0016] Fig. 1 is a block diagram showing an example of a configuration of an information processing device 100 according to the present embodiment. As shown in Fig. 1, the information processing device 100 includes a receiving unit 101, a learning unit 102, an output control unit 103, a storage unit 120, and a calculation unit 110.

[0017] The storage unit 120 stores various data used by the information processing device 100. For example, the storage unit 120 stores contract data 121, board information 122, learning data 123, a price estimation model 124, and a bid volume estimation model 125.

[0018] The storage unit 120 may be configured with any commonly used storage medium, such as a flash memory, a memory card, a RAM (Random Access Memory), a HDD (Hard Disk Drive), and an optical disk. The storage unit 120 may also be realized with a plurality of physically different storage media. For example, at least a portion of the plurality of data (contract data 121, order book information 122, learning data 123, price estimation model 124, and bid volume estimation model 125) shown in FIG. 1 may be stored in a distributed manner in a plurality of physically different storage media.

[0019] Fig. 2 is a diagram showing an example of the data structure of the contract data 121. The contract data 121 is data that represents a contract result in a spot market (an example of a first market). As shown in Fig. 2, the contract data 121 includes a spot market price, a spot market contract amount, and an area price.

[0020] The spot market price represents the price (agreement price) agreed upon in the spot market. The spot market price is equivalent to the system price, which represents the price agreed upon nationwide. The spot market agreement volume represents the trading volume of electricity agreed upon in the spot market. The area price represents the price agreed upon in each area divided into the whole country.

[0021] The data structure of the contract data 121 in Fig. 2 is an example and is not limited to this. Also, the "unit / format" in Fig. 2 is described to explain the unit or format of each element.

[0022] 3 is a diagram showing an example of the data structure of the depth of market information 122. The depth of market information 122 is data that represents the status of transactions in a pre-hours market (an example of a second market). As shown in FIG. 3, the depth of market information 122 mainly includes the following three types of information (data): Latest execution results ·Product information -Summary of transaction results

[0023] The latest contract result represents the result for the last executed bid in the open pre-hours market. In FIG. 3, the latest contract result includes three elements, but may also include other elements.

[0024] The commodity information is, for example, information for identifying a commodity (commodity ID). In the pre-hour market, for example, electricity traded in 48 periods obtained by dividing a day into 30-minute units is treated as a commodity. In this case, the commodity information may be a number (1 to 48) that can identify the 48 commodities (electricity).

[0025] The execution result summary represents information summarizing the results of bids that have already been executed in the open pre-hours market. In FIG. 3, the execution result summary includes 16 elements, but may include only some of these elements or may include other elements.

[0026] The contract data 121 and the depth information 122 are, for example, received by the receiving unit 101 and stored in the storage unit 120. The stored contract data 121 and depth information 122 are referenced in processing by the calculation unit 110, for example.

[0027] The learning data 123 is, for example, data (training data) used for learning (training) the price estimation model 124 and the bid volume estimation model 125. The learning data 123 of the price estimation model 124 is, for example, data including each element of the contract data 121 and each element of the order book information 122. In learning the price estimation model 124, a data set (learning data set) including a large number of learning data 123 is used.

[0028] The price estimation model 124 is a model that receives contract data 121 (including the trading volume of the trading object in the spot market) and order book information 122 as input data, and outputs the contract price of electricity in the ahead market. The price estimation model 124 is trained using training data 123 (training data set).

[0029] The price estimation model 124 can be interpreted as a machine learning model for estimating how the trading price of electricity in the hour-ahead market changes due to bidding behavior and the passage of time under the influence of market microstructure.

[0030] Fig. 5 is a diagram showing an overview of processing by the price estimation model 124. As shown in Fig. 5, the price estimation model 124 is a model that receives input data including depth information in the pre-hours market (corresponding to depth information 122) and contract results in the spot market (corresponding to contract data 121), and outputs a contract price that takes into account market price fluctuation factors as output data.

[0031] The bid amount estimation model 125 is a model that receives the contract data 121 and the estimated price as input data, and outputs the contract amount of electricity in the ahead market. The estimated price is a price selected from the contract price (first contract price) output by the price estimation model 124.

[0032] Returning to the description of Fig. 1, the receiving unit 101 receives input of various data used in the information processing device 100. For example, the receiving unit 101 receives input of contract data 121, order book information 122, and learning data 123 (learning data set).

[0033] The learning unit 102 executes a learning process for the price estimation model 124 and the bid volume estimation model 125. For example, the learning unit 102 learns the price estimation model 124 using learning data 123 including the contract data 121 (including the trading volume of the trading object in the spot market) and the order book information 122 as described above.

[0034] Here, a method for creating a learning data set used for learning will be described. The learning data set is created from order book information (corresponding to order book information 122) for all products in the time-ahead market within a certain period of time in the past, and contract results (corresponding to contract data 121) in the spot market.

[0035] Depth of market provides detailed information on all buy and sell bids for a particular slice of time. Depth of market shows the instantaneous supply and demand situation in the market and helps interpret market participants' behavior. Depth of market also includes time-ahead market execution results, such as product information, execution time, execution volume, and execution price.

[0036] The spot market contract results include the contract volume, system price, area price, etc., and reflect the overall market situation of the product. This information can be used to grasp the overall trend of the market.

[0037] The learning data set used for training the price estimation model 124 is created as a data set including a plurality of records including, for example, the same elements as the contract data 121 shown in Fig. 2 and the same elements as the depth of market information 122 shown in Fig. 3. In training the price estimation model 124, the latest contract price among the elements of the depth of market information 122 is used as a response variable (output of the model, correct answer data), and the other elements are used as explanatory variables (input of the model).

[0038] The learning data set used to train the bid volume estimation model 125 may be the same as or different from the learning data set used to train the price estimation model 124. Figure 4 is a diagram showing an example of a learning data set (learning data 123) used to train the bid volume estimation model 125 and different from the price estimation model 124.

[0039] In learning the bid volume estimation model 125, the latest contract volume among the elements of the board information 122 is used as a response variable (output of the model, correct answer data), and the other elements are used as explanatory variables (input of the model).

[0040] The learning unit 102 may perform pre-processing on at least a part of the elements included in the learning data set. The pre-processing is, for example, the following processing. Missing Values ​​Handling · Processing of outliers · Scaling the data (normalization or standardization) Encoding categorical data

[0041] Encoding categorical data is the process of dividing a contract date and time (an example of categorical data) expressed in the format year-month-day-date-hour-minute-second (yyyy-mm-dd hh:mm:ss) into six elements: year, month, day, hour, minute, and second.

[0042] Next, a description will be given of configuration examples of the price estimation model 124 and the bid volume estimation model 125. The price estimation model 124 and the bid volume estimation model 125 may be models of any structure, but are realized, for example, by a model using the following technique. Linear regression ·K nearest neighbors method Decision Tree Random Forest Gradient Boosting Neural Networks - Ensemble model combining multiple models above

[0043] The price estimation model 124 and the bid volume estimation model 125 are selected from the above models, taking into consideration the estimation accuracy, size, and the like.

[0044] The learning unit 102 learns the price estimation model 124 and the bid volume estimation model 125 using a learning data set. The learning method may be any conventionally used method, but for example, a method of adjusting the parameters of the model so as to minimize a loss function may be applied. The loss function is a value based on the difference between the observed value (correct answer data) and the predicted value (output data of the model). For example, in the case of a linear regression model, the mean square error or the like can be used as the loss function.

[0045] The technique for minimizing the loss function can be any conventional technique, such as gradient descent and variations of gradient descent.

[0046] The learning unit 102 may further tune hyperparameters to optimize the performance of each model. The hyperparameters are parameters that are set before learning, such as a learning rate, a depth of a decision tree, and a number of hidden layers. For tuning the hyperparameters, methods such as grid search, random search, and Bayesian optimization are used.

[0047] The learning unit 102 may perform cross-validation to prevent overfitting. For the cross-validation, a data set for validation (hereinafter, a validation data set) having the same format as the training data set is used. The learning unit 102 may be configured to use a part of the training data set as the validation data set.

[0048] For example, the training unit 102 trains each model using a training data set and evaluates (validates) the trained model using a validation data set. Cross-validation is the process of evaluating how a model performs on new data.

[0049] The learning unit 102 performs learning of each model in advance, for example, before the pre-hour market opens. Each trained model is stored in the storage unit 120. Note that the training data set may include training data corresponding to multiple products. Therefore, the pre-trained model can be interpreted as a model that can be commonly used for multiple products.

[0050] The learning unit 102 may further perform fine-tuning of each model to adapt it to the latest market conditions. Fine-tuning means fine-tuning the parameters of a pre-trained model by training it using another training data set.

[0051] For example, the learning unit 102 learns the price estimation model 124 using learning data including contract data 121 (including the trading volume of the trading object in the spot market) and depth information 122 obtained during a period DA (first period) from the opening of the pre-hour market to a certain time later. Fine tuning may be performed using learning data including depth information 122 for each product. This makes it possible to obtain a model that is finely tuned for each product.

[0052] Returning to the explanation of Fig. 1, the output control unit 103 controls the output of various information used in the information processing device 100. For example, the output control unit 103 outputs the bidding plan calculated by the calculation unit 110. Any output method may be used by the output control unit 103, and applicable methods include a method of displaying on a display device and a method of outputting via a network to an external device (such as a server device) that executes processing using the bidding plan.

[0053] The calculation unit 110 calculates a bidding plan including an optimal bid amount for electricity in the ahead market by using a price estimation model 124 and a bid amount estimation model 125. The calculation unit 110 includes a risk calculation unit 111, a bidable amount calculation unit 112, and a plan calculation unit 113.

[0054] The risk calculation unit 111 calculates an evaluation value (risk evaluation value) corresponding to one or more risk indexes. The calculated evaluation value is used when the plan calculation unit 113 calculates a bidding plan.

[0055] The risk indicators include, for example, the following indicators: (R1) Indicators based on the time until the pre-market close (time-based indicators) (R2) An indicator based on the difference between the maximum and maximum execution prices over a certain period of time (an indicator based on past data analysis) (R3) An index based on the difference between the trading volume of buy bids and the trading volume of sell bids (a bid imbalance-based index) (R4) An indicator based on the difference between the average contract price over a certain period and the latest contract price (price fluctuation-based indicator) (R5) An indicator based on the difference between the execution price of a buy bid and the execution price of a sell bid (spread-based indicator)

[0056] The function of the risk calculation unit 111 corresponds to evaluating the risk of price fluctuation and unsuccessful bids (imbalance) from real-time board information. For each risk index, multiple risk levels are set. In the following, an example will be described in which three risk levels (high, medium, low) are set. The risk levels are not limited to three levels, and for example, more detailed risk levels than three levels may be used.

[0057] The risk calculation unit 111 sets a specific criterion for each risk index, and evaluates the risk level of each risk index based on the set criterion. The risk calculation unit 111 does not need to use all of the above five risk indexes R1 to R5, and can be configured to use at least one risk index. For example, the risk calculation unit 111 may use only the time-based risk index R1 that can always be obtained.

[0058] Below, examples of risk assessment will be described for the five risk indexes R1 to R5.

[0059] (R1) Risk assessment using time-based indicators The time until the market closes is an effective factor in risk assessment. Normally, the risk is considered to increase in the period immediately before the market closes because the trends of market participants are uncertain. Therefore, the risk calculation unit 111 divides the time until the market closes into multiple periods and sets a risk level corresponding to each period.

[0060] For example, the risk calculation unit 111 sets three periods as the time until closing, and calculates a different risk assessment value for each period. For example, the period from 6 hours to 4 hours before closing is set as the first period. This period is determined to be a period of low risk (hereinafter, low risk period) because the time until closing is relatively long. Therefore, the risk calculation unit 111 calculates, for example, "2" as the risk assessment value if it corresponds to the low risk period. Next, the period from 4 hours to 1 hour before closing is set as the intermediate period. This period is determined to be a period of moderately high risk (hereinafter, medium risk period) because the time until closing is slightly shorter. Therefore, the risk calculation unit 111 calculates, for example, "1" as the risk assessment value if it corresponds to the medium risk period. Finally, the period within 1 hour until closing is set as the last period. This period is immediately before closing, and is determined to be a period with the highest risk (hereinafter, high risk period). Therefore, when the period corresponds to a high risk period, the risk calculation unit 111 calculates, for example, "0" as the risk evaluation value. The risk evaluation value quantified in this way makes it possible to quantitatively evaluate the risk level according to the time until closing. Note that the values ​​6 hours before, 4 hours before, and 1 hour before that are used to determine the three risk levels are merely examples, and other values ​​may be used.

[0061] (R2) Risk assessment using indicators based on historical data analysis The difference between the highest and lowest prices in the past 30 minutes (hereinafter, the price difference D_R2) reflects market volatility and is effective in understanding the risk of not reaching the target. The larger the price difference D_R2, the more unstable the market is judged to be and the higher the risk of not reaching the target is deemed to be. Therefore, in this embodiment, a method is adopted to quantify the price difference D_R2 in order to quantitatively evaluate the risk.

[0062] First, in order to divide the risk level into three stages, the price difference D_R2 is divided into three segments. For example, the price difference D_R2 is divided into three segments: 1 yen or less, 1 yen or more and less than 2 yen, and 2 yen or more. When the price difference D_R2 is 1 yen or less, the market is judged to be relatively stable and the risk is low. Therefore, the risk calculation unit 111 calculates a risk evaluation value of "2" when the segment corresponds to this. When the price difference D_R2 is 2 yen or more, the market is judged to be unstable and the risk is high. Therefore, the risk calculation unit 111 calculates a risk evaluation value of "0" when the segment corresponds to this. When the price difference D_R2 is more than 1 yen and less than 2 years, the market is judged to be a medium risk. Therefore, the risk calculation unit 111 calculates a risk evaluation value of "1" when the segment corresponds to this. Note that 1 yen and 2 yen, which are values ​​for determining the three stages of risk level, are only examples, and other values ​​may be used.

[0063] (R3) Risk assessment using bidding imbalance-based indicators The difference between the buy bid volume and sell bid volume in the board information (the value obtained by subtracting the sell bid volume from the buy bid volume; hereinafter, the difference D_R3) is an effective index for evaluating the risk of not achieving the contract price. For example, the larger the difference D_R3 is (the more the buy bid volume exceeds the sell bid volume), the lower the risk of selling electricity. Therefore, when the difference D_R3 is large, the risk is determined to be low.

[0064] For example, when the difference D_R3 is 50MWh or more, the market is judged to be relatively stable and the risk is low. In this case, the risk calculation unit 111 calculates "2" as the risk evaluation value. Conversely, when the difference D_R3 is negative (when the selling bid amount exceeds the buying bid amount), the risk of selling the power is high. Therefore, when the difference D_R3 is small, the risk is judged to be high. For example, when the difference D_R3 is -50MWh or less, the market is judged to be unstable and the risk is high. In this case, the risk calculation unit 111 calculates "0" as the risk evaluation value. In the case of an intermediate value, that is, when the difference D_R3 is greater than -50MWh and less than 50MWh, the risk is judged to be medium. In this case, the risk calculation unit 111 calculates "1" as the risk evaluation value. Note that the values ​​50MWh and -50MWh for determining the three-stage risk level are only examples, and other values ​​may be used. The above idea is an example of risk evaluation in the case of selling, but the same method can be applied to risk evaluation in the case of purchasing.

[0065] (R4) Risk assessment using price fluctuation-based indicators The absolute value of the difference between the weighted average price for the past 30 minutes and the latest contract price (hereinafter, the price difference D_R4) is an effective index representing the volatility of the market. The larger the absolute value of the price difference D_R4, the greater the fluctuation in market price and the higher the contract risk. For example, when the absolute value of the price difference D_R4 is 2 yen or more, the market is judged to have high volatility and high risk. In this case, the risk calculation unit 111 calculates "0" as the risk evaluation value. Conversely, when the absolute value of the price difference D_R4 is small, the market price is judged to have little fluctuation and low risk. For example, when the absolute value of the price difference D_R4 is 1 yen or less, the market is judged to have low volatility and low risk. In this case, the risk calculation unit 111 calculates "2" as the risk evaluation value. In the case of an intermediate value, that is, when the absolute value of the price difference D_R4 is greater than 1 yen and less than 2 yen, the market is judged to have medium volatility and medium risk. In this case, the risk calculation unit 111 calculates "1" as the risk evaluation value. Note that the values ​​of 1 yen and 2 yen used to define the three risk levels are merely examples, and other values ​​may be used.

[0066] (R5) Risk assessment using spread-based indices The absolute value of the price difference between the highest bid and the lowest ask in the board information (hereinafter, the price difference D_R5) is an index of liquidity and contract risk. If the absolute value of the price difference D_R5 is large, the risk that the market liquidity is low and contract is difficult increases. For example, if the absolute value of the price difference D_R5 is 2 yen or more, it is determined that the market liquidity is low and the risk is high. In this case, the risk calculation unit 111 calculates "0" as the risk evaluation value. Conversely, if the absolute value of the price difference D_R5 is small, it can be interpreted as a situation in which the market liquidity is high and contract is easy, so the risk is low. For example, if the absolute value of the price difference D_R5 is 1 yen or less, it is determined that the market liquidity is high and the risk is low. In this case, the risk calculation unit 111 calculates "2" as the risk evaluation value. In the case of an intermediate value, that is, if the absolute value of the price difference D_R5 is greater than 1 yen and less than 2 yen, it is determined that the market liquidity is medium and the risk is also medium. In this case, the risk calculation unit 111 calculates a risk evaluation value of "1." Note that the values ​​of 1 yen and 2 yen for determining the three risk levels are merely examples, and other values ​​may be used.

[0067] The risk calculation unit 111 sums up the risk evaluation values ​​calculated for each of the risk indexes used, and outputs the sum as a final risk evaluation value. The summed risk evaluation value represents the transaction risk of the market participants. For example, the larger the sum of the risk evaluation values, the lower the transaction risk is evaluated to be. By using multiple risk indexes, the accuracy of risk evaluation can be improved. Also, by evaluating each risk index in more detail, it is possible to increase the number of stages of risk evaluation values. In this case, risk evaluation that more accurately reflects the market situation is possible.

[0068] The biddable amount calculation unit 112 calculates the biddable amount, which is the amount of power that can be bid on the afore-hour market. In an example in which power is sold on the afore-hour market, the biddable amount corresponds to the amount of power that can be sold. For example, the biddable amount calculation unit 112 calculates the biddable amount from a predicted value of the amount of power generation and the amount of power traded in the past.

[0069] In addition, in the market, a minimum bidding unit (hereinafter, minimum bidding unit) may be determined. For example, in the hourly market, the minimum bidding unit is 50 kWh. In such a case, the biddable amount calculation unit 112 calculates the biddable amount to be a multiple of the minimum bidding unit.

[0070] The plan calculation unit 113 calculates a bidding plan for the afore-hours market by using the learned price estimation model 124, the risk evaluation value, and the available bid amount. The function of the plan calculation unit 113 corresponds to the above-mentioned functions F1 and F2.

[0071] For example, the plan calculation unit 113 calculates a plurality of contract prices P1 (first contract prices) corresponding to a plurality of bid volumes B1 (first bid volumes), respectively, using the trained price estimation model 124. The plurality of bid volumes B1 are calculated, for example, by sequentially adding up the minimum bidding unit of 50 kWh from the minimum bidding unit of 50 kWh to the biddable volume.

[0072] The plan calculation unit 113, for example, inputs input data including the board information 122 in which the bid volume B1 is set as the latest contract volume to the price estimation model 124, and can obtain the contract price output by the price estimation model 124 as the contract price P1.

[0073] When fine tuning of the price estimation model 124 is performed, the plan calculation unit 113 may calculate the contract price P1 using the fine-tuned price estimation model 124. In this case, the plan calculation unit 113 inputs input data based on the board information obtained during a period DB (second period) from the period DA during which the board information 122 used for fine tuning was obtained until the pre-hour market closes, to the price estimation model 124, to calculate a plurality of contract prices P1. The input data based on the board information is, for example, data in which the latest contract volume among the elements of the board information 122 obtained in the period DB is replaced with the bid volume B1.

[0074] Next, the plan calculation unit 113 selects a plurality of estimated prices EP1 representing the contract prices P1 corresponding to the plurality of risk levels from the plurality of contract prices P1. When three risk levels are used, the plan calculation unit 113 selects, for example, three contract prices P1 corresponding to three quartiles (25th percentile: first quartile, 50th percentile: median, 75th percentile: third quartile) of the plurality of bid volumes B1 as three estimated prices EP1 corresponding to the three risk levels. When two or four or more risk levels are used, the plan calculation unit 113 may select, for example, the contract price P1 corresponding to the position divided at a ratio corresponding to the number of levels as the estimated price EP1.

[0075] Next, the plan calculation unit 113 calculates a plurality of estimated bid volumes EB1 corresponding to the plurality of estimated prices EP1, respectively. A pair of the estimated price EP1 and the estimated bid volume EB1 corresponds to a candidate for a bidding plan. For example, the plan calculation unit 113 inputs input data including board information in which the estimated price EP1 is set as the latest contract price to the bid volume estimation model 125, and can obtain the contract volume output by the bid volume estimation model 125 as the estimated bid volume EB1.

[0076] The sum of the multiple estimated bid volumes EB1 may not match the available bid volume. Therefore, the plan calculation unit 113 may normalize the multiple estimated bid volumes EB1 corresponding to the multiple risk levels (e.g., three estimated bid volumes EB1 corresponding to three quartiles) so that the sum of the multiple estimated bid volumes EB1 corresponds to the available bid volume. For example, when using three estimated bid volumes EB1 corresponding to the quartiles, the plan calculation unit 113 performs normalization according to the following procedure. ·Let the available bid amount be X, and the three estimated bid amounts EB1 be a, b, and c. ·Calculate the sum of a, b, and c, S=a+b+c. The three estimated bid volumes EB1a, b, and c are normalized to obtain a', b', and c' using the following formula (1). a'=a×X / S,b'=b×X / S,c'=c×X / S...(1)

[0077] Bidding plan candidates may be calculated such that the biddable amount is distributed and allocated among a plurality of risk levels. When three risk levels are used, three bidding plan candidates are obtained. The three bidding plan candidates may be output as they are. For example, the output control unit 103 may output the calculated bidding plan candidates.

[0078] The plan calculation unit 113 may calculate one candidate selected from the bidding plan candidates as a recommended bidding plan. For example, the plan calculation unit 113 identifies a risk level L1 (first risk level) corresponding to the risk assessment value calculated by the risk calculation unit 111 from among a plurality of risk levels. When three risk levels and the above-mentioned five risk indexes R1 to R5 are used, the plan calculation unit 113 identifies the risk level L1 corresponding to the risk assessment value, for example, as follows. Note that the risk assessment value is the sum of the risk assessment values ​​(0, 1, or 2) for each of the five risk indexes, and takes a value from 0 to 10. - Risk assessment value is 0-3: Risk level is "high" - Risk assessment value is 4~6: Risk level is "medium" - Risk assessment value between 7 and 10: Risk level is "low"

[0079] When candidates for a bidding plan that distributes and allocates the biddable amount among multiple risk levels are calculated, and one candidate is calculated as the recommended bidding plan, a bidding plan that includes only a portion of the biddable amount is recommended. The remaining biddable amount is allocated in the bidding plan for the next or subsequent product. The plan calculation unit 113 calculates a bidding plan such that all the biddable amount is allocated by the time the market closes.

[0080] The plan calculation unit 113 calculates, as a recommended bidding plan, a bidding plan including a recommended price, which is the estimated price EP1 corresponding to the identified risk level L1, and a recommended bid quantity, which is the estimated bid quantity EB1 corresponding to the recommended price EP1.

[0081] FIG. 6 is a diagram showing an outline of the process by the calculation unit 110. As shown in FIG. 6, the calculation unit 110 inputs, as input data, the order book information (corresponding to order book information 122) of the pre-hour market and the contract result (corresponding to contract data 121) of the spot market, which are similar to the input data of the price estimation model 124, as well as calculation data such as the time until the closing, the time until the next calculation, and the available bid amount. Then, the calculation unit 110 outputs a plurality of bidding plan candidates as output data. Note that the time until the next calculation corresponds to the time until the bidding plan is recalculated. For example, 30 minutes, which corresponds to a product traded in the pre-hour market (electricity in 30-minute units), is input as the time until the next calculation.

[0082] At least a part of each of the above units (reception unit 101, learning unit 102, output control unit 103, and calculation unit 110) may be realized by one or more processing units. Each of the above units is realized, for example, by one or more processors. For example, each of the above units may be realized by having a processor such as a CPU (Central Processing Unit) and a GPU (Graphics Processing Unit) execute a program, that is, by software. Each of the above units may be realized by a processor such as a dedicated IC (Integrated Circuit), that is, by hardware. Each of the above units may be realized by using a combination of software and hardware. When multiple processors are used, each processor may realize one of the units, or two or more of the units.

[0083] The information processing device 100 may be physically configured as one device, or may be physically configured as multiple devices. For example, the information processing device 100 may be constructed on a cloud environment. Each unit in the information processing device 100 may be distributed and provided in multiple devices.

[0084] Next, a description will be given of a bidding plan calculation process performed by the information processing device 100 of this embodiment. Fig. 7 is a diagram showing an example of a method of operating the bidding plan calculation process.

[0085] 7, board information is accumulated for a certain period (e.g., from 17:00 to 23:00) after the opening of the pre-hour market, and the accumulated board information is used to learn (fine-tune) the price estimation model 124. After that (e.g., from 23:00 to 3:00 the next day), a bidding plan calculation process is executed using the learned price estimation model 124.

[0086] 8 is a flowchart showing an example of a bidding plan calculation process in this embodiment. Note that the flowchart in FIG. 8 includes a process of learning (fine-tuning) the price estimation model 124 as in FIG.

[0087] The reception unit 101 acquires, for each product, order book information for a certain period (e.g., from 5:00 p.m. to 11:00 p.m.) after the pre-hours market opens (step S101). The learning unit 102 uses the acquired order book information to train the price estimation model 124 (step S102).

[0088] Thereafter, the calculation of the bidding plan using the price estimation model 124 is performed in units of, for example, 30 minutes corresponding to one product.

[0089] The reception unit 101 acquires the latest board information (step S103). The latest board information corresponds to the board information of the currently targeted product. The risk calculation unit 111 calculates a risk evaluation value (step S104). The biddable amount calculation unit 112 calculates the biddable amount (step S105).

[0090] The plan calculation unit 113 calculates a plurality of contract prices P1 for a plurality of bid volumes B1 using the price estimation model 124 (step S106). The plan calculation unit 113 calculates an estimated bid volume EB1 corresponding to a plurality of estimated prices EP1 using the bid volume estimation model 125 (step S107). As described above, the estimated price EP1 is, for example, the contract price P1 corresponding to the three quartiles of the plurality of bid volumes B1.

[0091] The plan calculation unit 113 calculates bidding plan candidates including an estimated price EP1 and an estimated bid volume EB1. When a recommended bidding plan is to be found from the bidding plan candidates, the plan calculation unit 113 calculates the recommended bidding plan using the calculated risk evaluation value (step S108). The calculated bidding plan candidates and the recommended bidding plan are output by, for example, the output control unit 103.

[0092] In this way, in the information processing device of this embodiment, the contract price is estimated by a price estimation model that inputs the board information of the time-ahead market, and the more appropriate bidding plan is calculated from a plurality of bidding plans (bidding volume and bidding price) by referring to the risk assessment value of non-contraction calculated from the board information. This makes it possible to more accurately determine the appropriate bidding volume for a trading target in a market such as electricity.

[0093] Next, a hardware configuration of the information processing apparatus according to the embodiment will be described with reference to Fig. 9. Fig. 9 is an explanatory diagram illustrating an example of a hardware configuration of the information processing apparatus according to the embodiment.

[0094] The information processing device of the embodiment includes a control device such as a CPU (Central Processing Unit) 51, a storage device such as a ROM (Read Only Memory) 52 and a RAM (Random Access Memory) 53, a communication I / F 54 that connects to a network and communicates, and a bus 61 that connects each part.

[0095] The programs executed by the information processing apparatus according to the embodiment are provided in a state that they are pre-installed in the ROM 52 or the like.

[0096] The program executed by the information processing device of the embodiment may be configured to be provided as a computer program product by being recorded in an installable or executable format on a computer-readable recording medium such as a CD-ROM (Compact Disk Read Only Memory), a flexible disk (FD), a CD-R (Compact Disk Recordable), or a DVD (Digital Versatile Disk).

[0097] Furthermore, the program executed by the information processing apparatus of the embodiment may be configured to be stored on a computer connected to a network such as the Internet and provided by being downloaded via the network. Also, the program executed by the information processing apparatus of the embodiment may be configured to be provided or distributed via a network such as the Internet.

[0098] The programs executed by the information processing device of the embodiment can cause the computer to function as each unit of the information processing device described above. In this computer, the CPU 51 can read the programs from a computer-readable storage medium onto a main storage device and execute the programs.

[0099] A configuration example of the embodiment will be described below. (Configuration example 1) a trading volume of a trading object in a first market and order book information including a bid volume and a bid price of the trading object in a second market different from the first market are input, and a plurality of first contract prices corresponding to each of the plurality of first bid volumes are calculated using a price estimation model trained to output a contract price of the trading object in the second market; selecting a plurality of estimated prices representing the first contract prices respectively corresponding to a plurality of risk levels from the plurality of first contract prices; calculating a plurality of estimated bid quantities corresponding to the plurality of estimated prices; Processing section An information processing device comprising: (Configuration example 2) The processing unit includes: training the price estimation model using training data including the trading volume and the board information; The information processing device according to configuration example 1. (Configuration example 3) The processing unit includes: training the price estimation model using the learning data including the trading volume and the board information obtained during a first period from the opening of the second market until a certain time after the opening of the second market; inputting input data based on the board information obtained during a second period until the second market closes after the first period into the price estimation model, thereby calculating a plurality of the first contract prices; The information processing device according to configuration example 2. (Configuration Example 4) The processing unit includes: Calculating an assessment value corresponding to one or more risk indicators; Identifying a first risk level corresponding to the evaluation value among the plurality of risk levels; calculating a bidding plan including a recommended price, which is the estimated price corresponding to the first risk level among the plurality of estimated prices, and a recommended bid volume, which is the estimated bid volume corresponding to the recommended price among the plurality of estimated bid volumes; The information processing device according to any one of configuration examples 1 to 3. (Configuration Example 5) the risk indicator includes an indicator based on the time until the second market closes; The information processing device according to configuration example 4. (Configuration Example 6) The risk index is An indicator based on the difference between the maximum and maximum execution prices over a certain period of time; An indicator based on the difference between the volume of buying bids and the volume of selling bids; An indicator based on the difference between the average contract price over a certain period and the latest contract price, and An indicator based on the difference between the execution price of a buy bid and the execution price of a sell bid; Further comprising at least one of The information processing device according to configuration example 5. (Configuration Example 7) The processing unit includes: selecting three of the first contract prices corresponding to three quartiles of the first bid volumes as the estimated prices corresponding to the three risk levels; calculating a plurality of the estimated bid volumes corresponding to the respective plurality of the estimated prices using a bid volume estimation model that is trained to input the trading volume and the estimated price and output the contract volume of the trading object in the second market; The information processing device according to any one of configuration examples 1 to 6. (Configuration Example 8) The trading object is electricity, The trading amount is the amount of electricity sold or the amount of electricity purchased. The information processing device according to any one of configuration examples 1 to 7. (Configuration Example 9) The first market is a spot market; The second market is an ahead-of-time market. The information processing device according to configuration example 8. (Configuration Example 10) An information processing method executed by an information processing device, a step of inputting a trading volume of a trading object in a first market and order book information including a bid volume and a bid price of the trading object in a second market different from the first market, and calculating a plurality of first contract prices corresponding to the plurality of first bid volumes, respectively, using a price estimation model trained to output a contract price of the trading object in the second market; selecting, from the plurality of first contract prices, a plurality of estimated prices that represent the first contract prices respectively corresponding to a plurality of risk levels; calculating a plurality of estimated bid quantities corresponding to the plurality of estimated prices; An information processing method comprising: (Configuration Example 11) On the computer, a step of inputting a trading volume of a trading object in a first market and order book information including a bid volume and a bid price of the trading object in a second market different from the first market, and calculating a plurality of first contract prices corresponding to the plurality of first bid volumes, respectively, using a price estimation model trained to output a contract price of the trading object in the second market; selecting, from the plurality of first contract prices, a plurality of estimated prices that represent the first contract prices respectively corresponding to a plurality of risk levels; calculating a plurality of estimated bid quantities corresponding to the plurality of estimated prices; A program for executing.

[0100] Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These novel embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope and spirit of the invention, and are included in the scope of the invention and its equivalents described in the claims. [Explanation of symbols]

[0101] 100 Information processing device 101 Reception 102 Learning Department 103 Output control section 110 Calculation section 111 Risk Calculation Department 112 Bidding availability calculation unit 113 Planning Calculation Department 120 Storage section

Claims

1. a trading volume of a trading object in a first market and board information including a bid volume and a bid price of the trading object in a second market different from the first market are input, and a plurality of first contract prices corresponding to the plurality of first bid volumes are calculated using a price estimation model trained to output a contract price of the trading object in the second market; selecting a plurality of estimated prices representing the first contract prices respectively corresponding to a plurality of risk levels from the plurality of first contract prices; calculating a plurality of estimated bid quantities corresponding to the plurality of estimated prices; Processing section An information processing device comprising:

2. The processing unit includes: training the price estimation model using training data including the trading volume and the board information; The information processing device according to claim 1 .

3. The processing unit includes: training the price estimation model using the learning data including the trading volume and the board information obtained during a first period from the opening of the second market until a certain time after the opening of the second market; inputting input data based on the board information obtained during a second period until the second market closes after the first period into the price estimation model to calculate the first contract prices; The information processing device according to claim 2 .

4. The processing unit includes: Calculating an evaluation value corresponding to one or more risk indicators; Identifying a first risk level corresponding to the evaluation value among the plurality of risk levels; calculating a bidding plan including a recommended price, which is the estimated price corresponding to the first risk level among the plurality of estimated prices, and a recommended bid volume, which is the estimated bid volume corresponding to the recommended price among the plurality of estimated bid volumes; The information processing device according to claim 1 .

5. the risk indicator includes an indicator based on the time until the second market closes; The information processing device according to claim 4.

6. The risk index is An indicator based on the difference between the maximum and maximum execution prices over a certain period of time; An indicator based on the difference between the volume of buying bids and the volume of selling bids; An indicator based on the difference between the average contract price over a certain period and the latest contract price, and An indicator based on the difference between the execution price of a buy bid and the execution price of a sell bid; Further comprising at least one of: The information processing device according to claim 5 .

7. The processing unit includes: selecting three of the first contract prices corresponding to three quartiles of the first bid volumes as the estimated prices corresponding to the three risk levels; calculating a plurality of the estimated bid volumes corresponding to the respective plurality of the estimated prices using a bid volume estimation model that is trained to input the trading volume and the estimated price and output the contract volume of the trading object in the second market; The information processing device according to claim 1 .

8. The trading object is electricity, The trading amount is the amount of electricity sold or the amount of electricity purchased. The information processing device according to claim 1 .

9. the first market is a spot market; The second market is an ahead-of-time market. The information processing device according to claim 8.

10. An information processing method executed by an information processing device, a step of inputting a trading volume of a trading object in a first market and order book information including a bid volume and a bid price of the trading object in a second market different from the first market, and calculating a plurality of first contract prices corresponding to the plurality of first bid volumes, using a price estimation model trained to output a contract price of the trading object in the second market; selecting, from the plurality of first contract prices, a plurality of estimated prices representing the first contract prices respectively corresponding to a plurality of risk levels; calculating a plurality of estimated bid quantities corresponding to the plurality of estimated prices; An information processing method comprising:

11. On the computer, a step of inputting a trading volume of a trading object in a first market and order book information including a bid volume and a bid price of the trading object in a second market different from the first market, and calculating a plurality of first contract prices corresponding to the plurality of first bid volumes, using a price estimation model trained to output a contract price of the trading object in the second market; selecting, from the plurality of first contract prices, a plurality of estimated prices representing the first contract prices respectively corresponding to a plurality of risk levels; calculating a plurality of estimated bid quantities corresponding to the plurality of estimated prices; A program for executing the above.

Images