Sample construction method and device for spot market electricity price prediction based on feature correlation degree

Abstract

The application provides a spot market electricity price prediction sample construction method and device based on feature correlation, and belongs to the technical field of power system dispatching automation. The method acquires a historical sample library and operation feature data of a day to be predicted; key features are screened according to the correlation degree and correlation degree threshold value of various operation features and spot electricity prices, and the non-negative correlation degree weight corresponding to each key feature is determined according to the correlation degree corresponding to each key feature; the full-period feature distance of various key features is determined for each historical day, and is weighted and fused according to the non-negative correlation degree weight to obtain a feature correlation weighted distance; a similar day set is screened according to the feature correlation weighted distance and a distance threshold value, and an electricity price prediction sample set of the day to be predicted is constructed according to the similar day set. The application can reduce the interference of low correlation features on the similar day screening result.

Description

Technical Field

[0001] This invention belongs to the field of power system dispatch automation technology, and in particular relates to a method and apparatus for constructing spot market electricity price prediction samples based on feature correlation. Background Technology

[0002] As the development of the electricity market continues to deepen, electricity spot market prices are influenced by various factors, including supply and demand, fluctuations in renewable energy output, and inter-regional grid interactions. To predict future electricity price curves, existing technologies typically select historical days from historical data that exhibit similar operating trends to the day to be predicted, and use these similar days as sample data for price prediction. The selection of similar days usually employs a distance metric based on multi-dimensional operating characteristic sequences, comparing historical days with the day to be predicted in terms of system load forecasts, renewable energy output, tie-line load, and bidding space, and determining the similar days based on the distance.

[0003] However, the correlation between different spot market operating characteristics and spot electricity prices varies. Existing similar day screening methods typically use equal weighting to process various operating characteristics, and low-correlation or irrelevant characteristics are also included in distance calculations. This can easily lead to feature bias that masks characteristics with a high degree of correlation with spot electricity prices, resulting in a mismatch between the historical days selected and the actual electricity price change characteristics of the days to be predicted, thus reducing the quality of sample data used for spot market electricity price forecasting. Summary of the Invention

[0004] The purpose of this invention is to provide a method and apparatus for constructing spot market electricity price forecast samples based on feature correlation, so as to solve the problem that the existing technology uses an equal weighting method to process different spot market operating characteristics, resulting in low correlation features interfering with the screening results of similar days.

[0005] To achieve the above objectives, the following solution is adopted: In a first aspect, the present invention provides a method for constructing a sample for spot market electricity price forecasting based on feature correlation, comprising: Acquire historical sample database and operational characteristic data for the day to be predicted. The historical sample database includes multiple types of spot market operational characteristic data and spot electricity price data formed by time periods for multiple historical days. The operational characteristic data for the day to be predicted includes the multiple types of spot market operational characteristic data formed by time periods. Based on the various spot market operation characteristic data and spot electricity price data in the historical sample library, the correlation between various spot market operation characteristics and spot electricity prices is determined. The correlation between various spot market operation characteristics is compared with the correlation threshold to obtain a key feature set. Based on the correlation between each key feature in the key feature set, the non-negative correlation weight of each key feature is determined. For each historical day in the historical sample database, the feature distance of each type of key feature corresponding to the historical day is determined based on the feature value difference of the same type of key features between the historical day and the day to be predicted in each corresponding time period. Based on the non-negative correlation weights corresponding to each key feature, the feature distances of various key features corresponding to the historical day are weighted and fused across all time periods to obtain the feature correlation weighted distance between the day to be predicted and the historical day. Based on the feature correlation weighted distance and distance threshold between the date to be predicted and each historical day in the historical sample library, a set of similar days is obtained by filtering from the historical sample library. Then, based on the spot market operation feature data and spot electricity price data corresponding to each similar day in the set of similar days, a sample set for electricity price prediction of the date to be predicted is constructed.

[0006] The above scheme selects key features based on the correlation between the operating characteristics of various spot markets and spot electricity prices, and determines the non-negative correlation weights according to the correlation of each key feature. This allows key features with a high degree of correlation with spot electricity prices to play a greater role in historical day matching, reduces the interference of low-correlation features on the screening results of similar days, and thus improves the quality of sample data used for spot market electricity price forecasting.

[0007] Furthermore, the various spot market operating characteristics include system load forecasting, wind power forecasting, photovoltaic forecasting, tie-line power receiving load, total output of non-market-based units, and bidding space; The method further includes: before determining the correlation between various spot market operating characteristics and spot electricity prices and before determining the full-time characteristic distance of various key characteristics, normalizing the values ​​of various spot market operating characteristics in the historical sample library and the day to be predicted to obtain normalized operating characteristic values. The step of determining the correlation between various spot market operating characteristics and spot electricity prices based on the historical sample database of multiple types of spot market operating characteristic data and spot electricity price data includes: For each type of spot market operating characteristic, Pearson correlation analysis is used to calculate the correlation between the normalized operating characteristic value of that type of spot market operating characteristic and the spot electricity price in the historical sample database, so as to obtain the degree of correlation between that type of spot market operating characteristic and the spot electricity price.

[0008] The aforementioned further scheme selects multiple operating characteristics that can reflect the supply and demand status of the electricity spot market and normalizes the operating characteristic values ​​of different dimensions and numerical ranges to provide a unified data foundation for the subsequent determination of the characteristic distance throughout the entire time period. At the same time, Pearson correlation analysis is used to determine the degree of correlation between various operating characteristics and spot electricity prices, providing a basis for the selection of key characteristics and the determination of non-negative correlation weights.

[0009] Further, the step of comparing the correlation degree corresponding to various spot market operation characteristics with correlation thresholds to obtain a key feature set, and determining the non-negative correlation weight corresponding to each key feature based on the correlation degree corresponding to each key feature in the key feature set, includes: Spot market operation features whose absolute value of the corresponding correlation degree is less than the correlation degree threshold are removed, and the remaining spot market operation features are determined as key features to obtain the key feature set; Take the absolute value of the correlation degree corresponding to each key feature, and determine the non-negative correlation weight of the corresponding key feature based on each absolute value.

[0010] The above-mentioned further scheme removes operating features with low correlation to spot electricity prices by using a correlation threshold, and determines the non-negative correlation weights based on the absolute value of the correlation corresponding to each retained key feature. This ensures that the weights of each key feature can reflect the correlation strength between the corresponding key feature and the spot electricity price, and avoids the offsetting of positive and negative weights due to different correlation directions.

[0011] Further, determining the all-time feature distance of various key features corresponding to the historical day based on the feature value differences of the same key features between the historical day and the day to be predicted in each corresponding time period includes: For each type of key feature, calculate the Euclidean distance between the feature values ​​of the historical day and the day to be predicted in each corresponding time period to obtain the all-time feature distance of the key feature corresponding to the historical day. The step of weighted fusion of the full-time feature distances of various key features corresponding to the historical day based on the non-negative correlation weights of each key feature includes: Based on the non-negative correlation weights corresponding to each key feature, the feature distances of various key features corresponding to the historical day are weighted and summed over all time periods to obtain the feature correlation weighted distance between the day to be predicted and the historical day. The step of selecting a set of similar days from the historical sample database based on the weighted distance of feature correlation between the day to be predicted and each historical day in the historical sample database, and a distance threshold, includes: The historical days in the historical sample database are arranged in ascending order of the weighted distance of the feature relevance. Historical days whose corresponding feature relevance weighted distance is less than the distance threshold are identified as similar days, and the set of similar days is formed based on the identified similar days.

[0012] The above-mentioned further scheme uses Euclidean distance to represent the differences between historical days and the day to be predicted on the same key features throughout the entire time period, and performs weighted summation of the distances of various features throughout the entire time period according to the non-negative correlation weights corresponding to each key feature, so that the resulting feature correlation weighted distance can simultaneously reflect the differences of various key features throughout the entire time period and their correlation differences with spot electricity prices; through distance sorting and distance threshold screening, a set of similar days with a high degree of matching with the operating status of the day to be predicted can be formed.

[0013] In a second aspect, the present invention provides an apparatus for constructing a sample for spot market electricity price prediction based on feature correlation, comprising: The data acquisition module is used to acquire historical sample database and operational characteristic data of the day to be predicted. The historical sample database includes multiple types of spot market operational characteristic data and spot electricity price data formed by time periods for multiple historical days. The operational characteristic data of the day to be predicted includes the multiple types of spot market operational characteristic data formed by time periods. The feature processing module is used to determine the correlation between various spot market operation features and spot electricity prices based on the various spot market operation feature data and spot electricity price data in the historical sample library, compare the correlation between various spot market operation features with correlation thresholds to obtain a key feature set, and determine the non-negative correlation weight of each key feature based on the correlation between each key feature in the key feature set. The feature distance determination module is used to determine the full-time feature distance of various key features corresponding to each historical day in the historical sample library based on the feature value difference of the same key features between the historical day and the day to be predicted in each corresponding time period. The weighted fusion module is used to perform weighted fusion of the feature distances of various key features corresponding to the historical day throughout the entire time period according to the non-negative correlation weights of each key feature, so as to obtain the feature correlation weighted distance between the day to be predicted and the historical day. The sample construction module is used to select a set of similar days from the historical sample library based on the feature correlation weighted distance and distance threshold between the day to be predicted and each historical day in the historical sample library, and to construct a sample set of electricity price prediction for the day to be predicted based on the spot market operation feature data and spot electricity price data corresponding to each similar day in the set of similar days.

[0014] The aforementioned device filters key features and determines non-negative correlation weights through a feature processing module, obtains the feature correlation weighted distance for each historical day through a feature distance determination module and a weighted fusion module, and then forms a set of similar days based on a distance threshold using a sample construction module. This reduces the interference of low-correlation features on the similar day screening results and improves the adaptability of electricity price prediction samples.

[0015] Furthermore, the various spot market operating characteristics include system load forecasting, wind power forecasting, photovoltaic forecasting, tie-line power receiving load, total output of non-market-based units, and bidding space; The feature processing module is also used for: Before determining the correlation between various spot market operating characteristics and spot electricity prices, and before the feature distance determination module determines the full-time feature distance of various key characteristics, the values ​​of various spot market operating characteristics in the historical sample library and the day to be predicted are normalized to obtain normalized operating characteristic values. The feature processing module is specifically used to perform Pearson correlation analysis on each type of spot market operation feature to calculate the correlation between the normalized operation feature value of that type of spot market operation feature in the historical sample library and the spot electricity price, thereby obtaining the degree of correlation between that type of spot market operation feature and the spot electricity price.

[0016] The above-mentioned further scheme normalizes the operating characteristic values ​​of different dimensions and numerical ranges through the feature processing module, and uses Pearson correlation analysis to determine the correlation between various operating characteristics and spot electricity prices, providing a unified data foundation for key feature screening, weight determination and full-time feature distance calculation.

[0017] Furthermore, the feature processing module is specifically used for: Spot market operation features whose absolute value of the corresponding correlation degree is less than the correlation degree threshold are removed, and the remaining spot market operation features are determined as key features to obtain the key feature set; Take the absolute value of the correlation degree corresponding to each key feature, and determine the non-negative correlation weight of the corresponding key feature based on each absolute value.

[0018] The above-mentioned further scheme removes operational features with low correlation to spot electricity prices through a feature processing module, and determines the non-negative correlation weight based on the absolute value of the correlation corresponding to each key feature, so that the subsequent distance fusion can reflect the difference in the correlation strength between each key feature and the spot electricity price.

[0019] Furthermore, the feature distance determination module is specifically used to calculate the Euclidean distance between the feature values ​​of the historical day and the day to be predicted in each corresponding time period for each type of key feature, so as to obtain the full-time feature distance of the key feature corresponding to the historical day. The weighted fusion module is specifically used to perform a weighted summation of the feature distances of various key features corresponding to the historical day throughout the entire time period according to the non-negative correlation weights of each key feature, so as to obtain the feature correlation weighted distance between the day to be predicted and the historical day. The sample construction module is specifically used to arrange the historical days in the historical sample library in ascending order of the weighted distance of the feature relevance, determine the historical days whose corresponding weighted distance of the feature relevance is less than the distance threshold as similar days, and form the similar day set based on the determined similar days.

[0020] The above-mentioned further scheme calculates the Euclidean distances corresponding to various key features through the feature distance determination module, performs weighted summation according to non-negative correlation weights through the weighted fusion module, and sorts and filters the distances through the sample construction module, so that the resulting set of similar days can take into account the differences of various key features throughout the time period and their correlation differences with spot electricity prices.

[0021] In a third aspect, the present invention provides an electronic device including a processor and a memory, the processor being configured to execute a computer program stored in the memory to implement a method for constructing a spot market electricity price forecast sample based on feature correlation.

[0022] In a fourth aspect, the present invention provides a computer-readable storage medium storing at least one instruction that, when executed by a processor, implements a method for constructing a spot market electricity price prediction sample based on feature correlation. Attached Figure Description

[0023] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart illustrating a method for constructing a spot market electricity price forecast sample based on feature correlation, according to an embodiment of the present invention. Figure 2 This is a structural block diagram of a sample construction device for spot market electricity price prediction based on feature correlation, according to an embodiment of the present invention. Figure 3 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0024] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0025] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0026] Example 1 like Figure 1 As shown, this embodiment provides a method for constructing a spot market electricity price forecast sample based on feature correlation. This method selects historical days from a historical sample library that have similar operating trends to the day to be predicted, and uses the selected set of similar days to construct an electricity price forecast sample set for the day to be predicted, in order to forecast spot market electricity prices. This method considers the differences in the degree of correlation between different spot market operating characteristics and spot electricity prices. By eliminating low-correlation features and differentially assigning weights to the retained key features, key features with a high degree of correlation with spot electricity prices have higher weights during the historical day selection process. The method includes the following steps S1 to S5.

[0027] Step S1: Obtain the historical sample library and the operational characteristic data of the day to be predicted. The historical sample library includes multiple types of spot market operational characteristic data and spot electricity price data formed by time periods for multiple historical days. The operational characteristic data of the day to be predicted includes multiple types of spot market operational characteristic data formed by time periods.

[0028] Specifically, the historical sample database can consist of multi-source boundary data from the provincial electricity spot market. For each historical day in the historical sample database, the corresponding spot market operation characteristic data and spot electricity price data are organized according to the time periods included in the day, thus obtaining the daily dataset corresponding to each historical day. The operation characteristic data of the day to be predicted adopts the same feature type and time period organization method as the historical days, so as to compare the same characteristics of the day to be predicted with those of the historical days in the corresponding time periods in subsequent steps.

[0029] The operating characteristics of various spot markets include system load forecasting, wind power forecasting, photovoltaic forecasting, tie-line power receiving load, total output of non-market-based units, and bidding space.

[0030] Different types of spot market operating characteristics have different dimensions and numerical ranges. In order to enable various spot market operating characteristics to participate in subsequent correlation and feature distance calculations under a unified scale, it is necessary to normalize the corresponding feature values ​​to avoid features with larger numerical ranges masking features with smaller numerical ranges in distance calculations.

[0031] Before determining the correlation between various spot market operating characteristics and spot electricity prices, and before determining the full-time characteristic distances of various key characteristics, the operating characteristic values ​​of various spot markets in the historical sample database and the forecast date are normalized to obtain normalized operating characteristic values. Specifically, the Min-Max normalization method can be used, which involves subtracting the minimum value among the operating characteristic values ​​of each spot market from the minimum value of that category, and then dividing the difference by the difference between the maximum and minimum values ​​of that category of operating characteristic values. This is achieved through the following formula:

[0032] in, Represents the normalized operating characteristic value; This represents the characteristic value of the spot market operation before normalization; and These represent the maximum and minimum values ​​of the characteristic values ​​of this type of spot market operation, respectively.

[0033] Through the above processing, the operating characteristics of system load forecasting, wind power forecasting, photovoltaic forecasting, tie-line load, total output of non-market-based units, and bidding space are normalized to the same order of magnitude. Normalized operating characteristic values ​​from the historical sample database are used for correlation calculation in step S2 and for full-time characteristic distance calculation in step S3. The normalized operating characteristic value of the day to be predicted is used for comparison with historical days in step S3. All spot market operating characteristic values ​​in subsequent formulas are normalized spot market operating characteristic values.

[0034] Step S2: Based on the various spot market operation characteristic data and spot electricity price data in the historical sample library, determine the correlation between various spot market operation characteristics and spot electricity prices, compare the correlation corresponding to various spot market operation characteristics with the correlation threshold to obtain the key feature set, and determine the non-negative correlation weight corresponding to each key feature based on the correlation corresponding to each key feature in the key feature set.

[0035] The correlation between different spot market operating characteristics and spot electricity prices varies. For example, there is a strong correlation between bidding space and spot electricity prices, while the correlation between operating characteristics such as tie-line load and total output of non-market-based units and spot electricity prices may be weaker. If different operating characteristics are compared in historical daily data with equal weight, the bias of weakly correlated or irrelevant characteristics may mask the effect of strongly correlated characteristics. Therefore, this embodiment uses annual historical data to evaluate the correlation between various spot market operating characteristics and spot electricity prices, and selects key characteristics based on the evaluation results.

[0036] Determine the correlation between the operating characteristics of various spot markets and spot electricity prices, including the following processing: For each type of spot market operating characteristic, Pearson correlation analysis is used to calculate the correlation between the normalized operating characteristic value of that type of spot market operating characteristic and the spot electricity price in the historical sample database. Specifically, the electricity price deviation of the spot electricity price relative to the mean spot electricity price for each historical data point in the historical sample database, and the characteristic deviation of the normalized operating characteristic value of that type of spot market operating characteristic for each historical data point relative to the corresponding normalized operating characteristic mean are determined. The product of the electricity price deviation and the characteristic deviation for each historical data point is accumulated to obtain the deviation accumulation result. The square roots of the accumulated squares of the electricity price deviation and the accumulated squares of the characteristic deviation are taken respectively, and the product of the two square roots is used to normalize the deviation accumulation result to obtain the Pearson correlation coefficient between that type of spot market operating characteristic and the spot electricity price. This Pearson correlation coefficient is used as the degree of correlation between that type of spot market operating characteristic and the spot electricity price. That is, the Pearson correlation coefficient between the m-th type of spot market operating characteristic and the spot electricity price is calculated using the following formula:

[0037] in, Indicates the first The Pearson correlation coefficient between the operating characteristics of the spot market and spot electricity prices. Indicates the first Spot electricity prices at historical data points Indicates the first The first historical data point Normalized operating characteristic values ​​of spot market-like operating characteristics This represents the average spot electricity price across all historical data points. Represents the first historical data point The normalized mean of the operating characteristics of the spot market-like market. This indicates the number of historical data points.

[0038] In practice, for each type of spot market operating characteristic in the historical sample database, the Pearson correlation coefficient is calculated using the historical data for that type of spot market operating characteristic value and its corresponding spot electricity price throughout the year. The resulting... Used to characterize the The degree of linear correlation between the operating characteristics of the quasi-spot market and spot electricity prices is used as the basis for screening key characteristics and determining the weight of non-negative correlation.

[0039] The correlation between various spot market operating characteristics and correlation thresholds is compared, and a key feature set is determined based on the comparison results. The absolute value of the correlation corresponding to each key feature is taken, and the non-negative correlation weight of the corresponding key feature is determined based on each absolute value. Specifically, this includes: Spot market operating characteristics whose absolute values ​​of the corresponding Pearson correlation coefficients are less than the correlation threshold are removed, and the remaining spot market operating characteristics are identified as key characteristics, thus obtaining the key characteristic set. In one implementation, the correlation threshold can be set to 0.1.

[0040] The absolute value of the Pearson correlation coefficient is used to characterize the correlation strength between the corresponding operating characteristics and the spot electricity price. By comparing the absolute value of the Pearson correlation coefficient with 0.1, operating characteristics with low correlation strength can be eliminated as redundant features, so that subsequent historical daily comparisons are only performed on the key features in the key feature set.

[0041] In one implementation, the absolute value of the Pearson correlation coefficient corresponding to each key feature can be taken, and the non-negative correlation weight of the corresponding key feature can be determined according to the proportion of each absolute value in the sum of the absolute values ​​of the correlation of all key features. The specific formula is as follows:

[0042] in, Indicates the first The non-negative correlation weights corresponding to each key feature Indicates the first The Pearson correlation coefficient between key features and spot electricity prices Indicates the first The Pearson correlation coefficient between key features and spot electricity prices This indicates the number of key features in the key feature set.

[0043] This embodiment determines the non-negative correlation weights for each key feature by normalizing the absolute value of the Pearson correlation coefficient. The stronger the correlation between a key feature and the spot electricity price, the greater its corresponding non-negative correlation weight. Therefore, step S2 outputs a set of key features and the non-negative correlation weights for each key feature in the set. The key feature set is used to limit the feature comparison range in step S3, and the non-negative correlation weights are used for the all-time-period feature distance-weighted fusion in step S4.

[0044] Step S3: For each historical day in the historical sample library, determine the full-time feature distance of various key features corresponding to the historical day based on the feature value differences of the same key features between the historical day and the day to be predicted in each corresponding time period.

[0045] In practice, key feature values ​​of the day to be predicted and the current historical day can be extracted in a simulation comparison environment. For each type of key feature in the key feature set, the current historical day's [number] [value] [is then determined]. The characteristic values ​​of each time period and the predicted day The values ​​of similar features in each time period are compared to obtain the differences in the values ​​of similar key features in the corresponding time periods. Then, the differences in the values ​​of the key features of this type are summarized in all time periods of a single day to form the full-time feature distance of the key features of this type corresponding to the current historical day.

[0046] For each type of key feature, the Euclidean distance between the feature values ​​of the historical date and the date to be predicted in each corresponding time period is calculated to obtain the all-time feature distance of that type of key feature corresponding to that historical date. The specific calculation method is as follows: For the j-th historical day and the m-th key feature, the full-time feature distance of this type of key feature corresponding to that historical day can be calculated using the following formula:

[0047] in, Indicates the first The first historical day corresponds to the The full-time feature distance of each key feature. Indicates the first The first historical day The key feature in the first Feature values ​​for each time period Indicates the date to be predicted. The key feature in the first Feature values ​​for each time period This indicates the number of time periods included in a single day.

[0048] The above calculations are performed separately according to feature categories. After obtaining the... After determining the full-time feature distance corresponding to the first key feature, the same processing is performed on other key features in the key feature set, thereby obtaining the first key feature. Each historical day corresponds to The full-time feature distance of the key features is calculated, and the full-time feature distance is used for weighted fusion in step S4.

[0049] Step S4: Based on the non-negative correlation weights corresponding to each key feature, the feature distances of various key features corresponding to historical days are weighted and fused across all time periods to obtain the feature correlation weighted distance between the day to be predicted and historical days.

[0050] Based on the non-negative correlation weights corresponding to each key feature, the feature distances of various key features corresponding to historical days are weighted and summed over all time periods to obtain the feature correlation weighted distance between the day to be predicted and the historical day.

[0051] For the j-th historical day, the eigenvalue-weighted distance between the day to be predicted and that historical day can be calculated using the following formula:

[0052] in, Indicates the date to be predicted and the number Weighted distance of feature correlation between historical days Indicates the first The non-negative correlation weights corresponding to each key feature Indicates the first The first historical day corresponds to the The full-time feature distance of each key feature. This indicates the number of key features in the key feature set.

[0053] Key features with high relevance strength correspond to larger non-negative relevance weights, thus the feature distance of such key features across all time periods plays a significant role in the feature relevance weighted distance; key features with relatively low relevance strength but which are retained correspond to smaller non-negative relevance weights.

[0054] Feature relevance weighted distance is the comprehensive distance between the date to be predicted and the current historical date. The smaller this distance, the closer the changes of multiple key features in each time period of the current historical date are to those of the date to be predicted. By first determining the full-time feature distances of each type of key feature separately, and then weighting and fusing them according to relevance weights, a hierarchical calculation relationship can be formed for the feature value differences of corresponding time periods, the full-time feature distances of single-type key features, and the weighted distances of multiple types of key features.

[0055] After executing steps S1 and S2, steps S3 and S4 are executed for each historical day in the historical sample library to obtain the feature correlation weighted distance between the day to be predicted and each historical day in the historical sample library.

[0056] Steps S1 and S2 are used to form a uniform set of operational feature data, key features, and non-negative correlation weights for each key feature. Based on this, key feature values ​​for each historical day are read from the historical sample database one by one, and steps S3 and S4 are executed sequentially for the currently read historical day. After traversing all historical days in the historical sample database, the weighted distance of feature correlation for each historical day is obtained, which serves as the input for step S5 to sort historical days and filter similar days.

[0057] Step S5: Based on the feature correlation weighted distance and distance threshold between the day to be predicted and each historical day in the historical sample library, a set of similar days is selected from the historical sample library. Based on the spot market operation feature data and spot electricity price data corresponding to each similar day in the set of similar days, a sample set for electricity price prediction of the day to be predicted is constructed.

[0058] Specifically, the following processes are included: The historical days in the historical sample database are arranged in ascending order of feature relevance weighted distance; historical days whose corresponding feature relevance weighted distance is less than the distance threshold are identified as similar days, and a set of similar days is formed based on the identified similar days.

[0059] After traversing the historical sample database, the historical days ranked higher correspond to smaller feature relevance weighted distances, indicating a higher degree of similarity between these historical days and the day to be predicted. Similar days are extracted from the ranking results based on a distance threshold, and the resulting set of similar days is represented by the following formula:

[0060] in, This represents the generated set of similar days. Indicates the first in the historical sample database A historical day, Indicates the date to be predicted and the number Weighted distance of feature correlation between historical days This represents the distance threshold.

[0061] Therefore, the set of similar days This includes historical days in the historical sample library whose corresponding feature relevance weighted distance is less than a distance threshold. For each similar day in the set of similar days, the spot market operation characteristic data and spot electricity price data corresponding to the similar day, formed by time period, are used as a set of electricity price prediction sample data. The electricity price prediction sample data corresponding to each similar day are then aggregated to construct the electricity price prediction sample set for the day to be predicted.

[0062] This embodiment determines the correlation between various spot market operating characteristics and spot electricity prices using historical data throughout the year, eliminates low-correlation operating characteristics, and determines the non-negative correlation weights based on the correlation strength between the retained key characteristics and spot electricity prices. In the historical day comparison process, the full-time feature distances of various key characteristics are first formed, and then the full-time feature distances of various characteristics are weighted and fused to make the strongly correlated key characteristics play a greater role in historical day matching and reduce the interference of weakly correlated characteristics on the similar day screening results.

[0063] Example 2 like Figure 2 As shown, based on the same inventive concept as the above embodiments, the present invention also provides a sample construction device for spot market electricity price prediction based on feature correlation, including a data acquisition module, a feature processing module, a feature distance determination module, a weighted fusion module, and a sample construction module.

[0064] The data acquisition module is used to acquire historical sample databases and operational characteristic data for the day to be predicted. The historical sample database includes various types of spot market operational characteristic data and spot electricity price data formed by time periods for multiple historical days. The operational characteristic data for the day to be predicted includes various types of spot market operational characteristic data formed by time periods.

[0065] Specifically, the data acquisition module acquires various spot market operating characteristics, including system load forecasts, wind power forecasts, photovoltaic forecasts, tie-line loads, total output of non-market-based units, and bidding space. The historical sample database organizes corresponding operating characteristic data and spot electricity price data for each historical day according to daily time periods. The operating characteristic data for the day to be predicted adopts the same characteristic types and time periods as those for each historical day.

[0066] The feature processing module is also used to normalize the values ​​of various spot market operating characteristics in the historical sample library and the forecast date before determining the correlation between various spot market operating characteristics and spot electricity prices, and before the feature distance determination module determines the full-time feature distance of various key characteristics. This normalized operating characteristic value is then used for correlation calculation and full-time feature distance calculation. Specifically, the Min-Max normalization method described in Example 1 can be used.

[0067] The feature processing module is used to determine the correlation between various spot market operation features and spot electricity prices based on the historical sample database. It compares the correlation between various spot market operation features with correlation thresholds to obtain a key feature set, and determines the non-negative correlation weight of each key feature based on the correlation between each key feature in the key feature set.

[0068] Specifically, the feature processing module employs Pearson correlation analysis for each type of spot market operating characteristic to calculate the correlation between the normalized operating characteristic value of that type of spot market operating characteristic and the spot electricity price in the historical sample database, thereby obtaining the degree of correlation between the type of spot market operating characteristic and the spot electricity price. The feature processing module removes spot market operating characteristics whose absolute correlation value is less than a correlation threshold, and identifies the remaining spot market operating characteristics as key features, obtaining a key feature set. The correlation threshold is, for example, 0.1. Subsequently, the absolute value of the correlation degree corresponding to each key feature is taken, and the non-negative correlation weight of the corresponding key feature is determined based on each absolute value. In one implementation, the non-negative correlation weight can be determined according to the proportion of each absolute value in the sum of the absolute values ​​of the correlation degrees of all key features. The feature processing module provides the key feature set to the feature distance determination module and provides the non-negative correlation weights corresponding to each key feature to the weighted fusion module.

[0069] The feature distance determination module is used to determine the full-time feature distance of various key features corresponding to each historical day in the historical sample library based on the feature value differences of the same key features between the historical day and the day to be predicted in each corresponding time period.

[0070] Specifically, the feature distance determination module calculates the Euclidean distance between the feature values ​​of the historical day and the day to be predicted in each corresponding time period for each historical day and each type of key feature in the historical sample library, obtains the full-time feature distance of the key feature corresponding to the historical day, and provides each full-time feature distance to the weighted fusion module.

[0071] The weighted fusion module is used to perform weighted fusion of the feature distances of various key features corresponding to historical days across all time periods based on the non-negative correlation weights of each key feature, so as to obtain the weighted feature correlation distance between the day to be predicted and the historical days.

[0072] Specifically, the weighted fusion module performs a weighted sum of the feature distances of all types of key features corresponding to the current historical day according to the non-negative correlation weights of each key feature, to obtain the weighted feature correlation distance between the day to be predicted and the current historical day. The weighted fusion module performs the above processing on each historical day in the historical sample library to obtain the weighted feature correlation distance between the day to be predicted and each historical day, and provides the weighted feature correlation distance to the sample construction module.

[0073] The sample construction module is used to select a set of similar days from the historical sample library based on the feature correlation weighted distance and distance threshold between the day to be predicted and each historical day in the historical sample library, and to construct a sample set of electricity price prediction for the day to be predicted based on the spot market operation feature data and spot electricity price data corresponding to each similar day in the set of similar days.

[0074] Specifically, the sample construction module arranges the historical days in the historical sample database in ascending order of feature relevance weighted distance. Historical days with a feature relevance weighted distance less than a distance threshold are identified as similar days, and a set of similar days is formed based on these similar days. For each similar day in the set, the corresponding time-period spot market operation characteristic data and spot electricity price data are used as a set of electricity price prediction sample data. These electricity price prediction sample data for each similar day are then aggregated to construct the electricity price prediction sample set for the day to be predicted.

[0075] The apparatus in this embodiment is based on the same inventive concept as the method in Embodiment 1. The specific processing methods and corresponding technical effects of each module can be found in the corresponding descriptions in Embodiment 1, and will not be repeated here.

[0076] Example 3 like Figure 3As shown, the present invention also provides an electronic device 100 for implementing a method for constructing a sample for spot market electricity price prediction based on feature correlation; The electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104.

[0077] The memory 101 can be used to store the computer program 103. The processor 102 implements the sample construction method for spot market electricity price prediction based on feature correlation in Embodiment 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0078] The memory 101 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0079] At least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 102 may be a microprocessor or any conventional processor. Processor 102 is the control center of electronic device 100, connecting various parts of electronic device 100 via various interfaces and lines.

[0080] The memory 101 in the electronic device 100 stores multiple instructions to implement a method for constructing a spot market electricity price forecast sample based on feature correlation, and the processor 102 can execute multiple instructions to achieve the following: Acquire historical sample database and operational characteristic data for the day to be predicted. The historical sample database includes multiple types of spot market operational characteristic data and spot electricity price data formed by time periods for multiple historical days. The operational characteristic data for the day to be predicted includes the multiple types of spot market operational characteristic data formed by time periods. Based on the various spot market operation characteristic data and spot electricity price data in the historical sample library, the correlation between various spot market operation characteristics and spot electricity prices is determined. The correlation between various spot market operation characteristics is compared with the correlation threshold to obtain a key feature set. Based on the correlation between each key feature in the key feature set, the non-negative correlation weight of each key feature is determined. For each historical day in the historical sample database, the feature distance of each type of key feature corresponding to the historical day is determined based on the feature value difference of the same type of key features between the historical day and the day to be predicted in each corresponding time period. Based on the non-negative correlation weights corresponding to each key feature, the feature distances of various key features corresponding to the historical day are weighted and fused across all time periods to obtain the feature correlation weighted distance between the day to be predicted and the historical day. Based on the feature correlation weighted distance and distance threshold between the date to be predicted and each historical day in the historical sample library, a set of similar days is obtained by filtering from the historical sample library. Then, based on the spot market operation feature data and spot electricity price data corresponding to each similar day in the set of similar days, a sample set for electricity price prediction of the date to be predicted is constructed.

[0081] Example 4 If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, and read-only memory (ROM).

[0082] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0083] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0084] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0085] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0086] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0087] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for constructing a sample for spot market electricity price forecasting based on feature correlation, characterized in that, include: Acquire historical sample database and operational characteristic data for the day to be predicted. The historical sample database includes multiple types of spot market operational characteristic data and spot electricity price data formed by time periods for multiple historical days. The operational characteristic data for the day to be predicted includes the multiple types of spot market operational characteristic data formed by time periods. Based on the various spot market operation characteristic data and spot electricity price data in the historical sample library, the correlation between various spot market operation characteristics and spot electricity prices is determined. The correlation between various spot market operation characteristics is compared with the correlation threshold to obtain a key feature set. Based on the correlation between each key feature in the key feature set, the non-negative correlation weight of each key feature is determined. For each historical day in the historical sample database, the feature distance of each type of key feature corresponding to the historical day is determined based on the feature value difference of the same type of key features between the historical day and the day to be predicted in each corresponding time period. Based on the non-negative correlation weights corresponding to each key feature, the feature distances of various key features corresponding to the historical day are weighted and fused across all time periods to obtain the feature correlation weighted distance between the day to be predicted and the historical day. Based on the feature correlation weighted distance and distance threshold between the date to be predicted and each historical day in the historical sample library, a set of similar days is obtained by filtering from the historical sample library. Then, based on the spot market operation feature data and spot electricity price data corresponding to each similar day in the set of similar days, a sample set for electricity price prediction of the date to be predicted is constructed.

2. The method for constructing a sample for spot market electricity price forecasting based on feature correlation according to claim 1, characterized in that, The various spot market operating characteristics include system load forecasting, wind power forecasting, photovoltaic forecasting, tie-line power receiving load, total output of non-market-based units, and bidding space; The method further includes: Before determining the correlation between various spot market operating characteristics and spot electricity prices, and before determining the full-time characteristic distance of various key characteristics, the values ​​of various spot market operating characteristics in the historical sample library and the day to be predicted are normalized to obtain normalized operating characteristic values. The step of determining the correlation between various spot market operating characteristics and spot electricity prices based on the historical sample database of multiple types of spot market operating characteristic data and spot electricity price data includes: For each type of spot market operating characteristic, Pearson correlation analysis is used to calculate the correlation between the normalized operating characteristic value of that type of spot market operating characteristic and the spot electricity price in the historical sample database, so as to obtain the degree of correlation between that type of spot market operating characteristic and the spot electricity price.

3. The method for constructing a sample for spot market electricity price forecasting based on feature correlation according to claim 1, characterized in that, The process of comparing the correlation degree corresponding to various spot market operation characteristics with correlation thresholds to obtain a key feature set, and determining the non-negative correlation weight corresponding to each key feature based on the correlation degree corresponding to each key feature in the key feature set, includes: Spot market operation features whose absolute value of the corresponding correlation degree is less than the correlation degree threshold are removed, and the remaining spot market operation features are determined as key features to obtain the key feature set; Take the absolute value of the correlation degree corresponding to each key feature, and determine the non-negative correlation weight of the corresponding key feature based on each absolute value.

4. The method for constructing a sample for spot market electricity price forecasting based on feature correlation according to claim 1, characterized in that, The step of determining the all-time feature distance of various key features corresponding to the historical day based on the feature value differences of the same key features between the historical day and the day to be predicted in each corresponding time period includes: For each type of key feature, calculate the Euclidean distance between the feature values ​​of the historical day and the day to be predicted in each corresponding time period to obtain the all-time feature distance of the key feature corresponding to the historical day. The step of weighted fusion of the full-time feature distances of various key features corresponding to the historical day based on the non-negative correlation weights of each key feature includes: Based on the non-negative correlation weights corresponding to each key feature, the feature distances of various key features corresponding to the historical day are weighted and summed over all time periods to obtain the feature correlation weighted distance between the day to be predicted and the historical day. The step of selecting a set of similar days from the historical sample database based on the weighted distance of feature correlation between the day to be predicted and each historical day in the historical sample database, and a distance threshold, includes: The historical days in the historical sample database are arranged in ascending order of the weighted distance of the feature relevance. Historical days whose corresponding feature relevance weighted distance is less than the distance threshold are identified as similar days, and the set of similar days is formed based on the identified similar days.

5. A sample construction device for spot market electricity price forecasting based on feature correlation, characterized in that, include: The data acquisition module is used to acquire historical sample database and operational characteristic data of the day to be predicted. The historical sample database includes multiple types of spot market operational characteristic data and spot electricity price data formed by time periods for multiple historical days. The operational characteristic data of the day to be predicted includes the multiple types of spot market operational characteristic data formed by time periods. The feature processing module is used to determine the correlation between various spot market operation features and spot electricity prices based on the various spot market operation feature data and spot electricity price data in the historical sample library, compare the correlation between various spot market operation features with correlation thresholds to obtain a key feature set, and determine the non-negative correlation weight of each key feature based on the correlation between each key feature in the key feature set. The feature distance determination module is used to determine the full-time feature distance of various key features corresponding to each historical day in the historical sample library based on the feature value difference of the same key features between the historical day and the day to be predicted in each corresponding time period. The weighted fusion module is used to perform weighted fusion of the feature distances of various key features corresponding to the historical day throughout the entire time period according to the non-negative correlation weights of each key feature, so as to obtain the feature correlation weighted distance between the day to be predicted and the historical day. The sample construction module is used to select a set of similar days from the historical sample library based on the feature correlation weighted distance and distance threshold between the day to be predicted and each historical day in the historical sample library, and to construct a sample set of electricity price prediction for the day to be predicted based on the spot market operation feature data and spot electricity price data corresponding to each similar day in the set of similar days.

6. The sample construction device for spot market electricity price prediction based on feature correlation according to claim 5, characterized in that, The various spot market operating characteristics include system load forecasting, wind power forecasting, photovoltaic forecasting, tie-line power receiving load, total output of non-market-based units, and bidding space; The feature processing module is further configured to: before determining the correlation between various spot market operating characteristics and spot electricity prices and before the feature distance determination module determines the full-time feature distance of various key characteristics, normalize the values ​​of various spot market operating characteristics in the historical sample library and the day to be predicted to obtain normalized operating characteristic values. The feature processing module is specifically used to perform Pearson correlation analysis on each type of spot market operation feature to calculate the correlation between the normalized operation feature value of that type of spot market operation feature in the historical sample library and the spot electricity price, thereby obtaining the degree of correlation between that type of spot market operation feature and the spot electricity price.

7. The sample construction device for spot market electricity price prediction based on feature correlation according to claim 5, characterized in that, The feature processing module is specifically used for: Spot market operation features whose absolute value of the corresponding correlation degree is less than the correlation degree threshold are removed, and the remaining spot market operation features are determined as key features to obtain the key feature set; Take the absolute value of the correlation degree corresponding to each key feature, and determine the non-negative correlation weight of the corresponding key feature based on each absolute value.

8. The sample construction device for spot market electricity price prediction based on feature correlation according to claim 5, characterized in that: The feature distance determination module is specifically used to calculate the Euclidean distance between the feature values ​​of the historical day and the day to be predicted in each corresponding time period for each type of key feature, so as to obtain the full-time feature distance of the key feature corresponding to the historical day. The weighted fusion module is specifically used to perform a weighted summation of the feature distances of various key features corresponding to the historical day throughout the entire time period according to the non-negative correlation weights of each key feature, so as to obtain the feature correlation weighted distance between the day to be predicted and the historical day. The sample construction module is specifically used to arrange the historical days in the historical sample library in ascending order of the weighted distance of the feature relevance, determine the historical days whose corresponding weighted distance of the feature relevance is less than the distance threshold as similar days, and form the similar day set based on the determined similar days.

9. An electronic device, characterized in that, It includes a processor and a memory, the processor being used to execute a computer program stored in the memory to implement the method for constructing spot market electricity price forecast samples based on feature correlation as described in any one of claims 1 to 4.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction, which, when executed by a processor, implements the method for constructing a spot market electricity price forecast sample based on feature correlation as described in any one of claims 1 to 4.

Images