Power market price prediction system

Abstract

To provide a power market price prediction system that can easily maintain the prediction accuracy of power market prices.SOLUTION: A power market price prediction system 100 predicts power market prices. When it is determined that there is a deviation between a prediction value and an actual value of the power market prices that is equal to or more than a predetermined threshold (YES in step S13), a prediction model creating unit creates a first prediction model by learning actual value information including a deviating actual value, and creates a second prediction model by learning the actual value information excluding the deviating actual value (step S14). A price prediction unit obtains a first prediction value of the power market prices using the first prediction model, obtains a second prediction value of the power market prices using the second prediction model, and outputs an average value of a plurality of prediction values including at least the first prediction value and the second prediction value as a prediction value of the power market prices (step S15).SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present invention relates to a technology for an electricity market price forecasting system for forecasting electricity market prices. [Background technology]

[0002] Conventionally, technology for an electricity market price prediction system that predicts electricity market prices has been publicly known, as described in Patent Document 1, for example.

[0003] Patent Document 1 discloses an electricity market price prediction device that predicts the electricity market price based on an estimated amount of electricity supplied by renewable energy on a target prediction date, actual value information on the amount of electricity supplied, and actual value information on the electricity market price. In this way, the technology described in Patent Document 1 aims to improve the accuracy of electricity market price prediction by predicting the estimated amount of electricity supplied and the amount of electricity demanded, which have a large impact on the electricity market price. By improving the accuracy of electricity market price prediction, it is possible to improve the revenue from selling electricity.

[0004] Here, if a sudden event such as a breakdown in power generation facilities occurs, the event may cause a sudden price fluctuation (price spike) in the electricity market price. In the technology described in Patent Document 1, in order to reduce prediction errors in the electricity market price, the price spike portion is excluded from the actual value information of the electricity market price.

[0005] However, if the sudden price fluctuation is not irregular (it will continue in the future), excluding the sudden price fluctuation from the actual value information of the electricity market price may result in an inability to maintain the accuracy of the electricity market price forecast. However, if the sudden price fluctuation is included in the actual value information of the electricity market price, there is a problem that the error in the electricity market price forecast will increase if the sudden price fluctuation is irregular (it will not continue in the future). Whether the sudden price fluctuation is irregular or not can be determined after a few days based on external information and price trends, but it is desirable to easily maintain the accuracy of the electricity market price forecast until then. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Publication No. 2019-96164 Summary of the Invention [Problem to be solved by the invention]

[0007] The present invention has been made in consideration of the above-mentioned circumstances, and the problem that it aims to solve is to provide an electricity market price prediction system that can easily maintain the prediction accuracy of electricity market prices. [Means for solving the problem]

[0008] The problem to be solved by the present invention is as described above, and the means for solving this problem will now be described.

[0009] That is, claim 1 provides an electricity market price prediction system for predicting an electricity market price, comprising: an actual value information acquisition unit that acquires actual value information including past actual values ​​of the electricity market price; a prediction model creation unit that creates a prediction model for predicting the electricity market price by learning the actual value information; and a price prediction unit that acquires a predicted value of the electricity market price using the prediction model, wherein when the prediction model creation unit determines that there is a deviation between the predicted value and the actual value of the electricity market price that is equal to or greater than a predetermined threshold, the prediction model creation unit creates a first prediction model by learning the actual value information including the deviating actual value, and creates a second prediction model by learning the actual value information excluding the deviating actual value, and the price prediction unit acquires a first predicted value of the electricity market price using the first prediction model, acquires a second predicted value of the electricity market price using the second prediction model, and sets an average value of a plurality of predicted values ​​including at least the first predicted value and the second predicted value as the predicted value of the electricity market price.

[0010] In claim 2, the plurality of predicted values ​​includes an average value of the electricity market price for a recent predetermined period.

[0011] In claim 3, the plurality of forecast values ​​includes the electricity market price of the previous day.

[0012] In claim 4, the prediction model creation unit creates the second prediction model by learning the performance value information in which the removed deviated performance values ​​are complemented with past performance values.

[0013] In claim 5, the predetermined threshold is set to a value that differs depending on the time period.

[0014] In claim 6, after the price prediction unit outputs the average value of the multiple predicted values ​​as the predicted value of the electricity market price, if the prediction model creation unit determines that the deviating actual value is irregular, it re-learns the actual value information excluding the deviating actual value, and if the prediction model creation unit determines that the deviating actual value is not irregular, it re-learns the actual value information including the deviating actual value. [Effects of the Invention]

[0015] The present invention has the following effects.

[0016] In the invention according to claim 1, the prediction accuracy of the electricity market price can be easily maintained.

[0017] In the invention according to claim 2, it is possible to improve the accuracy of prediction of the electricity market price.

[0018] In the invention according to claim 3, it is possible to improve the accuracy of prediction of the electricity market price.

[0019] In the invention according to claim 4, it is possible to improve the accuracy of prediction of the electricity market price.

[0020] In the invention according to claim 5, it is possible to appropriately judge the accuracy of the prediction of the electricity market price.

[0021] In the invention according to claim 6, it is possible to improve the accuracy of prediction of the electricity market price. [Brief explanation of the drawings]

[0022] [Figure 1] 1 is a block diagram showing the configuration of an electricity market price prediction system according to an embodiment of the present invention. [Figure 2] 1 is a flowchart illustrating predictive control for predicting electricity market prices. [Figure 3](a) A graph showing an example of actual electricity market prices. (b) A graph showing actual electricity market prices with rapid price fluctuations removed. [Figure 4] 10 is a flowchart showing the control of relearning determination of actual value information of the electricity market price. DETAILED DESCRIPTION OF THE INVENTION

[0023] An electricity market price prediction system 100 according to one embodiment of the present invention will be described below.

[0024] An electricity market price prediction system 100 according to one embodiment of the present invention predicts electricity market prices in the wholesale electricity market (hereinafter simply referred to as "electricity market prices"). As shown in Fig. 1, the electricity market price prediction system 100 includes an acquisition unit 110 and a calculation unit 120.

[0025] The acquisition unit 110 acquires various types of information. The acquisition unit 110 can acquire information for predicting the electricity market price, etc., via a network such as the Internet. The information for predicting the electricity market price includes, for example, temperature, solar radiation, wind speed, electricity demand, and fuel price. The information for predicting the electricity market price also includes actual value information including past actual values ​​of the electricity market price.

[0026] The calculation unit 120 performs various calculations. The calculation unit 120 can predict the electricity market price based on the information acquired by the acquisition unit 110. Specifically, the calculation unit 120 can acquire actual value information including past actual values ​​of the electricity market price, and create a prediction model for predicting the electricity market price by learning the acquired actual value information.

[0027] Here, a "prediction model" is a model that includes machine learning, with the electricity market price as the "objective variable" and temperature, solar radiation, wind speed, electricity demand, fuel price, etc. as "explanatory variables." For example, in the case of a multiple regression model, a regression coefficient, which is the coefficient of the explanatory variable, is set for each explanatory variable (temperature, solar radiation, wind speed, electricity demand, fuel price, etc.). The regression coefficient indicates the contribution rate (weight) of each explanatory variable to the electricity market price (objective variable). If the regression coefficient is positive, an increase in the explanatory variable will increase the electricity market price (objective variable), and if the regression coefficient is negative, an increase in the explanatory variable will decrease the electricity market price (objective variable).

[0028] The calculation unit 120 can predict the electricity market price using the created prediction model and output the predicted electricity market price. The predicted value of the electricity market price is obtained every predetermined time unit, for example, every 48 frames obtained by dividing a day into 30-minute units.

[0029] The electricity market price prediction system 100 configured in this manner obtains the predicted electricity market price results before bidding in the wholesale electricity market, allowing users to predict the bidding curve (supply and demand curve) for bidding in the wholesale electricity market.

[0030] The main wholesale electricity markets are the day-ahead market (spot market) and the intraday market (hour-ahead market). The day-ahead market is a market where electricity is traded the day before actual supply and demand. In the day-ahead market, trading takes place for 48 products (48 frames), dividing the day into 30-minute increments. On the other hand, the intraday market is a market where trading takes place up to one hour before actual supply and demand, and is mainly responsible for adjusting surpluses and shortages in the day-ahead market. In the following, the electricity market price prediction system 100 according to this embodiment will be described as predicting wholesale electricity market prices in the day-ahead market (spot market).

[0031] Here, when predicting the electricity market price, there may be a large deviation between the predicted value and the actual value of the electricity market price. If an actual value that deviates significantly from the predicted value (hereinafter referred to as the "deviation part") is trained to create a prediction model, and this deviation part is irregular (will not continue in the future), the prediction error of the electricity market price will become large. On the other hand, if the deviation part is not irregular (will continue in the future), the prediction accuracy of the electricity market price cannot be maintained unless the deviation part is also trained. Whether the deviation part is irregular or not can be determined after a few days based on external information and price trends, but it is desirable to maintain the prediction accuracy of the electricity market price until then.

[0032] Therefore, in the electricity market price prediction system 100 according to this embodiment, the calculation unit 120 predicts the electricity market price by executing the prediction control shown in Fig. 2. Hereinafter, a method for predicting the electricity market price will be described with reference to the flowchart shown in Fig. 2.

[0033] In the following, the day on which bidding in the wholesale electricity market is conducted will be referred to as "the day," the day of actual supply and demand (i.e., the day following the day on which bidding in the day-ahead market is conducted) will be referred to as "the next day," and the day after that will be referred to as "the day after that." In the wholesale electricity market, bidding must be conducted by 10:00 AM on the day. In the prediction control shown in FIG. 2, the processing of steps S11 and S12 is for predicting the electricity market price on the next day. The processing of steps S11 and S12 is performed by 10:00 AM on the day, which is the bidding time in the wholesale electricity market, for example, at 6:00 AM on the same day. Furthermore, the processing of steps S13 to S16 is for predicting the electricity market price on the day after that. The processing of steps S13 to S16 is performed by 10:00 AM on the next day, for example, at 6:00 AM on the next day.

[0034] In step S11, the calculation unit 120 executes a calculation process (a process of predicting the electricity market price) using a first prediction model. The "first prediction model" will be described later.

[0035] After performing the process of step S11, the calculation unit 120 proceeds to step S12.

[0036] In step S12, the calculation unit 120 outputs the predicted value (prediction result) of the electricity market price for the next day calculated by the calculation process in step S11.

[0037] After performing the process of step S12, the calculation unit 120 proceeds to step S13.

[0038] At 10:00 AM on the day, which is the bidding time in the wholesale electricity market, the final value of the electricity market price for the next day (hereinafter referred to as the "actual value") becomes clear.

[0039] In step S13, the calculation unit 120 determines whether the prediction error is greater than A (prediction error > A). Here, the "prediction error" indicates the difference between the predicted value of the electricity market price output in step S12 and the actual value of the electricity market price. In this embodiment, the prediction error is indicated by an absolute value. Furthermore, "A" is a predetermined threshold value that is set based on the actual values ​​of the electricity market price in the past, etc.

[0040] The calculation unit 120 compares the predicted value and actual value of the electricity market price for each time slot (each of the 48 frames) of the day and calculates the prediction error for each time slot (frame). If there is one or more time slots among the time slots (48 frames) of the day where the prediction error is greater than A, the calculation unit 120 determines YES in step S13.

[0041] Fig. 3(a) shows an example of the transition of the actual value of the electricity market price. In the example shown in Fig. 3(a), a sudden price fluctuation occurs on June 25th. When such a sudden price fluctuation occurs, there is a high possibility that the predicted value and the actual value of the electricity market price will diverge significantly (a divergence will occur), and the prediction error will be greater than A ("YES" in step S13).

[0042] If the calculation unit 120 determines that the prediction error is greater than A ("YES" in step S13), the process proceeds to step S14. On the other hand, if the calculation unit 120 determines that the prediction error is not greater than A ("NO" in step S13), the process proceeds to step S16.

[0043] In step S14, the calculation unit 120 creates a first prediction model and a second prediction model. Here, the "first prediction model" is a prediction model created by learning actual value information including the deviation portion (the portion where the prediction error is greater than A). Also, the "second prediction model" is a prediction model created by learning actual value information excluding the deviation portion (the portion where the prediction error is greater than A) (without learning the deviation portion).

[0044] In the example shown in Fig. 3, if the sudden price fluctuation on June 25th results in a forecast error > A ("YES" in step S13), the calculation unit 120 will train the sudden price fluctuation (deviation portion) and then create a first forecast model based on the performance value information including the deviation portion. Alternatively, the calculation unit 120 will not train the sudden price fluctuation (deviation portion) and will create a second forecast model based on the performance value information excluding the deviation portion. The "performance value information excluding the deviation portion" may be performance value information excluding only the deviation portion, or it may be performance value information excluding the deviation portion and the performance values ​​for the time period before and after the deviation portion (for example, the one day (48 frames) when the deviation occurred).

[0045] For example, as shown in FIG. 3(b), the calculation unit 120 excludes the actual values ​​for the time period in which the deviation occurred or the actual values ​​for one day (48 frames) in which the deviation occurred from the actual value information of the electricity market price. Then, the calculation unit 120 creates a second prediction model by learning the actual value information from which the deviation has been excluded. The excluded part (the part not learned) may be supplemented with past actual values ​​(the average value of the electricity market price for the most recent predetermined period (e.g., seven days), the electricity market price of the previous day, etc.). This makes it possible to maintain the continuity of the actual values.

[0046] After performing the process of step S14, the calculation unit 120 proceeds to step S15.

[0047] In step S15, the calculation unit 120 outputs the average value of the predicted value using the first prediction model and the predicted value using the second prediction model as the prediction result of the electricity market price for the day after next. Specifically, the calculation unit 120 obtains the predicted value of the electricity market price for the day after next using the first prediction model. Furthermore, the calculation unit 120 obtains the predicted value of the electricity market price for the day after next using the second prediction model. Then, the calculation unit 120 calculates the average value of the predicted value obtained using the first prediction model (first predicted value) and the predicted value obtained using the second prediction model (second predicted value), and outputs the calculated average value as the prediction result of the electricity market price for the day after next.

[0048] After performing the process of step S15, the calculation unit 120 ends the predictive control shown in FIG.

[0049] Meanwhile, in step S16, the calculation unit 120 executes a calculation process using the first prediction model and outputs the prediction result of the electricity market price for the day after that (the day after the day on which step S16 is executed).

[0050] After performing the process of step S16, the calculation unit 120 ends the predictive control shown in FIG.

[0051] As described above, in the electricity market price prediction system 100 according to this embodiment, when there is a large discrepancy between the predicted value and the actual value of the electricity market price, a predicted value using a first prediction model created after learning the discrepancy portion (the portion where the prediction error is > A) and a predicted value using a second prediction model created without learning the discrepancy portion are obtained, and the average value of the predicted value using the first prediction model and the predicted value using the second prediction model is used as the prediction result for the electricity market price. This makes it possible to prevent the error in the prediction result from becoming large, regardless of whether the discrepancy portion is irregular, for several days until it becomes clear whether the discrepancy portion is irregular or not.

[0052] Next, the re-learning determination control will be explained using Fig. 4. As described above, when there is a large deviation between the predicted value and the actual value of the electricity market price, the average value of the predicted value using the first prediction model and the predicted value using the second prediction model is used as the prediction result of the electricity market price, but the re-learning determination control in Fig. 4 determines whether the deviation was irregular several days later, and depending on the result, determines whether to re-learn the deviation.

[0053] Hereinafter, the day on which there is a large deviation between the predicted value and the actual measured value (a deviation portion occurs) will be referred to as the "first day," the day after the first day will be referred to as the "second day," the day after the second day will be referred to as the "third day," and the day after the third day will be referred to as the "fourth day." In the re-learning determination control shown in FIG. 4, the processes of steps S21 and S22 are executed on the third day, and the processes of steps S23 and S26 are executed on the fourth day. Furthermore, if the answer is YES in step S22, the processes of steps S24 and S25 are executed on the third day. On the other hand, if the answer is YES in step S23, the processes of steps S24 and S25 are executed on the fourth day.

[0054] In step S21, the calculation unit 120 acquires data for relearning determination. For example, the calculation unit 120 acquires information on the actual measured value of the electricity market price on the first day, when there was a large discrepancy between the predicted value and the actual measured value, and information on the actual value of the electricity market price from the second day to the third day.

[0055] After performing the process of step S21, the calculation unit 120 proceeds to step S22.

[0056] In step S22, the calculation unit 120 determines whether the price trend has returned to normal. Specifically, the calculation unit 120 determines whether the electricity market price on the third day (particularly the electricity market price during the time period when the deviation occurred on the first day) has approached the electricity market price before the first day. Any method can be used to determine whether the price trend has returned to normal.

[0057] For example, the calculation unit 120 compares the actual value of the electricity market price on the first day with the actual value of the electricity market price on the third day, and makes the determination based on the comparison result. More specifically, the calculation unit 120 determines that the price trend has returned if the electricity market price during the time period in which the deviation occurred on the first day changes by a predetermined value (e.g., 5 yen / kWh) or more on the third day in a direction returning to the original state (approaching the electricity market price before the first day). For example, if the price fluctuated sharply downward (toward a lower electricity market price) on the first day and the electricity market price increased by 5 yen / kWh or more on the third day, the calculation unit 120 determines that the price trend has returned. Furthermore, if the price fluctuated sharply upward (toward a higher electricity market price) on the first day and the electricity market price decreased by 5 yen / kWh or more on the third day, the calculation unit 120 determines that the price trend has returned.

[0058] If the calculation unit 120 determines that the price trend has reversed ("YES" in step S22), the process proceeds to step S24. On the other hand, if the calculation unit 120 determines that the price trend has not reversed ("NO" in step S22), the process proceeds to step S23.

[0059] In step S23, the calculation unit 120 determines whether the price trend has returned. Specifically, the calculation unit 120 compares the actual value of the electricity market price on the first day with the actual value of the electricity market price on the fourth day, and makes this determination based on the comparison result. For example, the calculation unit 120 determines that the price trend has returned if the electricity market price during the time period in which the divergence occurred on the first day has changed by more than a predetermined value (e.g., 5 yen / kWh) in the direction toward the original price on the fourth day.

[0060] If the calculation unit 120 determines that the price trend has reversed ("YES" in step S23), the process proceeds to step S24. On the other hand, if the calculation unit 120 determines that the price trend has not reversed ("NO" in step S23), the process proceeds to step S26.

[0061] In step S24, the calculation unit 120 determines that the deviation of the actual value of the electricity market price on the first day from the predicted value is irregular.

[0062] After performing the process of step S24, the calculation unit 120 proceeds to step S25.

[0063] In step S25, the calculation unit 120 re-learns the actual value information of the electricity market price excluding the deviation portion on day 1, and updates the actual value information. Note that the calculation unit 120 may re-learn excluding only the deviation portion, or may re-learn excluding the deviation portion and the actual values ​​for the time period before and after the deviation portion (for example, the one day (48 frames) when the deviation occurred).

[0064] After performing the process of step S25, the calculation unit 120 ends the flow of FIG.

[0065] Meanwhile, in step S26, the calculation unit 120 re-learns the actual value information of the electricity market price, including the deviation portion on the first day, and updates the actual value information.

[0066] After performing the process of step S26, the calculation unit 120 ends the re-learning determination control shown in FIG.

[0067] As described above, in the electricity market price prediction system 100 according to this embodiment, if the electricity market price during the time period in which the deviation occurred returns to the previous trend of the electricity market price, the deviation is determined to be irregular. If the deviation is determined to be irregular, the deviation is excluded from the actual value information of the electricity market price, re-learning is performed, and the actual value information is updated. If the deviation is determined not to be irregular, the deviation is included in the actual value information of the electricity market price, learning is performed, and the actual value information is updated. In this way, by increasing the reliability of the actual value information of the electricity market price, it is possible to improve the prediction accuracy of the electricity market price.

[0068] As described above, the electricity market price prediction system 100 according to this embodiment: An electricity market price forecasting system 100 for forecasting an electricity market price, a performance value information acquisition unit (acquisition unit 110, calculation unit 120) that acquires performance value information including past performance values ​​of the electricity market price; a prediction model creation unit (calculation unit 120) that creates a prediction model for predicting the electricity market price by learning the actual value information; a price prediction unit (calculation unit 120) that obtains a predicted value of the electricity market price using the prediction model; Equipped with The prediction model creation unit If it is determined that there is a deviation of a predetermined threshold or more between the predicted value and the actual value of the electricity market price (YES in step S13 of FIG. 2), a first prediction model is created by learning the actual value information including the deviated actual value, and a second prediction model is created by learning the actual value information excluding the deviated actual value (step S14). The price prediction unit obtaining a first forecast of the electricity market price using the first forecast model; obtaining a second forecast of the electricity market price using the second forecasting model; An average value of a plurality of predicted values ​​including at least the first predicted value and the second predicted value is set as the predicted value of the electricity market price (step S15).

[0069] With this configuration, the prediction accuracy of the electricity market price can be easily maintained. Specifically, for several days until it becomes clear whether the deviating actual value (deviation portion) is irregular, it is possible to prevent the error in the forecast result from increasing, regardless of whether the deviation portion is irregular or not. In addition, it is possible to easily obtain the forecast result of the electricity market price without obtaining external information.

[0070] Further, the prediction model creation unit The second prediction model is created by learning the performance value information in which the removed deviated performance values ​​are complemented with past performance values ​​(step S14).

[0071] This configuration can improve the accuracy of forecasting the electricity market price. Specifically, it is possible to maintain the continuity of actual values ​​of the electricity market price, and thus to improve the prediction accuracy of the electricity market price.

[0072] Further, the prediction model creation unit After the price prediction unit outputs the average value of the plurality of prediction values ​​as the prediction value of the electricity market price (step S15 in FIG. 2), If it is determined that the deviated performance value is irregular (YES in step S22 or YES in step S23 in FIG. 4), the performance value information excluding the deviated performance value is re-learned (step S25). If it is determined that the deviated performance value is not irregular, the performance value information including the deviated performance value is re-learned (step S26).

[0073] This configuration can improve the accuracy of forecasting the electricity market price. Specifically, by determining whether or not to include the deviating actual value (deviation portion) in the actual value information of the electricity market price depending on whether the deviation portion is irregular, the reliability of the actual value information of the electricity market price can be increased, and ultimately the prediction accuracy of the electricity market price can be improved.

[0074] Although the embodiment of the present invention has been described above, the present invention is not limited to the above configuration, and various modifications are possible within the scope of the invention described in the claims.

[0075] For example, in this embodiment, the prediction model is based on multiple regression analysis, but it may also be based on other methods such as random forest.

[0076] In this embodiment, the "forecast error" in step S13 of Fig. 2 is expressed as an absolute value representing the difference between the predicted value and the actual value of the electricity market price. However, the forecast error may be expressed as a positive or negative value instead of an absolute value. Alternatively, the forecast error may be expressed as a ratio (deviation rate) of the difference between the predicted value and the actual value of the electricity market price to the actual value of the electricity market price. In this case, the threshold A is also set to a value corresponding to the deviation rate.

[0077] Furthermore, in this embodiment, when the predicted value and the actual value of the electricity market price significantly deviate from each other ("YES" in step S13), the calculation unit 120 outputs the average value of the predicted value using the first prediction model (first predicted value) and the predicted value using the second prediction model (second predicted value) as the prediction result of the electricity market price. However, the average value of the electricity market price for the most recent predetermined period (e.g., seven days) or the electricity market price of the previous day may also be used to calculate the average value. For example, the calculation unit 120 may output the average value of the average electricity market price for the most recent predetermined period and the first predicted value and the second predicted value as the prediction result of the electricity market price. Alternatively, the calculation unit 120 may output the average value of the electricity market price for the previous day and the first predicted value and the second predicted value as the prediction result of the electricity market price. Alternatively, the calculation unit 120 may output the average value of the electricity market price for the most recent predetermined period and the average value of the electricity market price of the previous day and the first predicted value and the second predicted value as the prediction result of the electricity market price. In this way, by calculating an average value using actual values ​​other than the first predicted value and the second predicted value (average value of electricity market prices for the most recent specified period (e.g., seven days), electricity market price of the previous day), and outputting the calculated average value as the predicted result of the electricity market price, the accuracy of the electricity market price prediction can be improved.

[0078] As described above, in the electricity market price prediction system 100 according to this embodiment, The plurality of predicted values ​​include: The average value of the electricity market price for the most recent predetermined period is included.

[0079] This configuration can improve the accuracy of forecasting the electricity market price. in particular,

[0080] The plurality of predicted values ​​may include: The electricity market price of the previous day is included.

[0081] This configuration can improve the accuracy of forecasting the electricity market price.

[0082] 2 may be set individually for each time period (for each of the 48 frames). Also, threshold A may be set individually for each season (summer, winter, intermediate season). Since the trend of the electricity market price varies depending on the time period and season, by setting threshold A individually for each time period and season, it is possible to appropriately determine the accuracy of the electricity market price prediction in step S13.

[0083] As described above, in the electricity market price prediction system 100 according to this embodiment, The predetermined threshold value is It is set to different values ​​depending on the time of day and / or season.

[0084] With this configuration, it is possible to appropriately determine whether the accuracy of the electricity market price prediction is good or bad.

[0085] Furthermore, in this embodiment, the calculation unit 120 outputs the average value of the predicted value (first predicted value) obtained using the first prediction model and the predicted value (second predicted value) obtained using the second prediction model as the predicted result of the electricity market price for the day after the next (step S15 in Figure 2), but the average value may be an arithmetic average or another type of average value such as a weighted average.

[0086] Furthermore, in the re-learning judgment control of Figure 4, if the electricity market price during the time period in which the deviation occurred on the first day changes back to the original price by, for example, 5 yen / kWh or more on a later day, it is determined that the price trend has returned (steps S22, S23), but it may also be determined that the price trend has returned when the prediction error falls below threshold A.

[0087] In step S22 of Fig. 4, the calculation unit 120 performs the irregularity determination (determines whether the price trend has returned) based on the comparison result between the actual value of the electricity market price on the first day and the actual value of the electricity market price on the third day. However, the calculation unit 120 may also perform the irregularity determination by taking into account the actual value of the electricity market price on the second day. That is, the calculation unit 120 may determine that the price trend has returned if the electricity market price in the time slot where the deviation occurred on the first day changes by a predetermined value (e.g., 5 yen / kWh) or more on each of the second and third days (both the second and third days) in the direction of returning to the original state (in the direction of approaching the electricity market price before the first day). Similarly, in step S23 of Fig. 4, the calculation unit 120 may also perform the irregularity determination by taking into account the actual value of the electricity market price on the second day.

[0088] Furthermore, if multiple sudden price fluctuations in different directions (upward and downward) occur during the execution of the re-learning judgment in Figure 4 (for example, if a sudden price fluctuation in the upward direction (toward higher prices) occurs on the first day and a sudden price fluctuation in the downward direction (toward lower prices) occurs on the second day), a re-learning judgment may be performed for each of them.

[0089] Furthermore, the processing performed by the calculation unit 120 does not necessarily have to be performed by the calculation unit 120 alone, but may be performed in cooperation with (outsourced to) a calculation unit of another system, for example. [Explanation of symbols]

[0090] 100 Electricity Market Price Forecasting System 110 Acquisition Department 120 Arithmetic section

Claims

1. An electricity market price prediction system for predicting an electricity market price, a performance value information acquiring unit that acquires performance value information including past performance values ​​of the electricity market price; a prediction model creation unit that creates a prediction model for predicting the electricity market price by learning the actual value information; a price prediction unit that obtains a predicted value of the electricity market price using the prediction model; Equipped with The prediction model creation unit When it is determined that there is a deviation of a predetermined threshold or more between the predicted value and the actual value of the electricity market price, a first prediction model is created by learning the actual value information including the deviated actual value, and a second prediction model is created by learning the actual value information excluding the deviated actual value; The price prediction unit obtaining a first forecast of the electricity market price using the first forecast model; obtaining a second forecast of the electricity market price using the second forecasting model; an average value of a plurality of forecast values ​​including at least the first forecast value and the second forecast value is set as the forecast value of the electricity market price; Electricity market price forecasting system.

2. The plurality of predicted values ​​include: The average value of the electricity market price for a recent predetermined period is included. The electricity market price forecasting system of claim 1 .

3. The plurality of predicted values ​​include: The electricity market price for the previous day is included. The electricity market price forecasting system according to claim 1 or 2.

4. The prediction model creation unit creating the second prediction model by learning the performance value information obtained by complementing the removed deviated performance values ​​with past performance values; The electricity market price forecasting system of claim 1 .

5. The predetermined threshold value is It is set to different values ​​depending on the time. The electricity market price forecasting system of claim 1 .

6. The prediction model creation unit After the price prediction unit outputs an average value of the plurality of predicted values ​​as a predicted value of the electricity market price, If it is determined that the deviated performance value is irregular, the performance value information excluding the deviated performance value is re-learned; If it is determined that the deviated performance value is not irregular, the performance value information is re-learned including the deviated performance value. The electricity market price forecasting system of claim 1 .

Images