Medium and long term electricity price prediction method and device, electronic equipment and storage medium

Abstract

The invention provides a medium and long term electricity price prediction method and device, electronic equipment and a storage medium, and belongs to the technical field of data prediction, and the method comprises the steps: determining a first feature set based on the correlation between a multi-dimensional market boundary condition feature and a market clearing electricity price, and carrying out the data processing of the first feature set, and obtaining a second feature set; training the initial electricity price prediction model based on the second feature set to obtain a target electricity price prediction model; and on the basis of the historical data of the first time before the to-be-measured time point and the target electricity price prediction model, obtaining an electricity price prediction result of the second time after the to-be-measured time point. According to the medium-and-long-term electricity price prediction method and device, the electronic equipment and the storage medium provided by the invention, the accuracy of medium-and-long-term electricity price prediction can be improved.

Description

Technical Field

[0001] The present disclosure belongs to the technical field of data prediction, and more specifically, relates to a medium- and long-term electricity price prediction method and device, electronic equipment, and storage medium. Background Art

[0002] Electricity price forecast is to estimate and infer the price of electricity market in the future. Accurate electricity price forecast can reflect the supply and demand situation of electricity market, provide decision-making basis for power dispatching agencies, and make power resources more reasonably allocated in different regions and different time periods.

[0003] The current electricity price forecast is mainly aimed at short-term electricity price forecasts and cannot meet actual needs. A small number of forecasts are made for medium- and long-term electricity prices, but there are still shortcomings. For example, the traditional machine learning algorithm used cannot make predictions when the feature value is missing; the time series method used can only be used to deduce a single sequence and cannot take into account other related features. It is not only inaccurate but also cannot support long-term forecasts. Since it is impossible to obtain daily electricity price data and related conditional characteristics in the future long-term period, existing technologies often lack consideration of multi-dimensional feature correlation and long-term time series continuity, and the resulting prediction value is of low accuracy. Summary of the invention

[0004] The purpose of the present disclosure is to provide a medium- and long-term electricity price forecasting method and device, electronic equipment, and storage medium to improve the accuracy of medium- and long-term electricity price forecasting.

[0005] A first aspect of the embodiments of the present disclosure provides a method for predicting medium- and long-term electricity prices, including: A first feature set is determined based on the correlation between the multi-dimensional market boundary condition characteristics and the market clearing electricity price, and a second feature set is obtained by performing data processing on the first feature set.

[0006] The initial electricity price prediction model is trained based on the second feature set to obtain a target electricity price prediction model.

[0007] The electricity price prediction result at the second time after the time point to be tested is obtained based on the historical data of the first time before the time point to be tested and the target electricity price prediction model.

[0008] A second aspect of the embodiments of the present disclosure provides a medium- and long-term electricity price prediction device, comprising: The feature extraction module is used to determine a first feature set based on the correlation between the multi-dimensional market boundary condition characteristics and the market clearing electricity price, and to perform data processing on the first feature set to obtain a second feature set.

[0009] The model training module is used to train the initial electricity price prediction model based on the second feature set to obtain a target electricity price prediction model.

[0010] The electricity price prediction module is used to obtain the electricity price prediction result at the second time after the time point to be tested based on the historical data at the first time before the time point to be tested and the target electricity price prediction model.

[0011] According to a third aspect of an embodiment of the present disclosure, an electronic device is provided, including a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the steps of the above-mentioned medium- and long-term electricity price forecasting method when executing the computer program.

[0012] According to a fourth aspect of the embodiments of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned method for medium- and long-term electricity price forecasting are implemented.

[0013] The beneficial effects of a medium- and long-term electricity price prediction method and device, electronic device, and storage medium provided by the embodiments of the present disclosure are: On the one hand, the embodiment of the present disclosure deeply analyzes the correlation between the multi-dimensional market boundary condition characteristics and the market clearing electricity price, selects the characteristics with strong correlation and performs data processing. This step effectively integrates the multi-dimensional factors affecting the electricity price, lays a solid foundation for the subsequent electricity price prediction model training, reduces the interference of irrelevant information, and makes the prediction results more accurate.

[0014] On the other hand, the embodiment of the present disclosure uses the processed feature set data to train the initial electricity price prediction model, which can fully capture the complex laws of market supply and demand changes, thereby generating a high-precision target electricity price prediction model. The target electricity price prediction model not only considers the continuity of the time series, but also incorporates the correlation between multi-dimensional features, significantly improving the comprehensiveness and accuracy of the prediction.

[0015] On the other hand, based on the historical data before the test time point and the trained target electricity price prediction model, the electricity price data within a longer time period can be reliably predicted, providing strong decision-making support for power dispatching agencies, and helping to achieve a more reasonable allocation of power resources and smooth operation of the market. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A flowchart of a medium- and long-term electricity price prediction method provided by an embodiment of the present disclosure; Figure 2 A schematic diagram of the structure of a CRFNet deep neural network provided in an embodiment of the present disclosure; Figure 3 A structural block diagram of a medium- and long-term electricity price prediction device provided by an embodiment of the present disclosure; Figure 4 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION

[0018] In the following description, specific details such as specific system structures and technologies are provided for the purpose of illustration rather than limitation, so as to provide a thorough understanding of the embodiments of the present disclosure. However, it should be clear to those skilled in the art that the present disclosure may be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present disclosure with unnecessary details.

[0019] In order to make the purpose, technical solutions and advantages of the present disclosure more clear, specific embodiments will be described below in conjunction with the accompanying drawings.

[0020] Please refer to Figure 1 , Figure 1 A flowchart of a medium- and long-term electricity price forecasting method provided in an embodiment of the present disclosure, the method may include S101 to S103.

[0021] S101: Determine a first feature set based on the correlation between the multi-dimensional market boundary condition characteristics and the market clearing electricity price, and perform data processing on the first feature set to obtain a second feature set.

[0022] In this embodiment, the multidimensional market boundary condition characteristics may include load characteristics and environmental characteristics. Load characteristics may include thermal power load value, new energy load value, provincial dispatch load value, external power load value, non-market output load value, bidding space, hydropower load, photovoltaic load, wind power load, etc. Environmental characteristics may include weather type, temperature, etc. The first feature set includes multiple features that are screened out from the multidimensional market boundary condition characteristics and have a strong correlation with the market clearing electricity price (the correlation coefficient is greater than the first threshold).

[0023] In this embodiment, data processing is performed on the first feature set to obtain the second feature set, including: performing outlier screening and missing value completion on the first feature set to obtain the second feature set.

[0024] Exemplarily, the quartile method is used to detect outliers in the original input data (i.e., the second feature set). The quartile method includes three parameters, and the original input data is divided into four parts according to these three parameters. The three parameters are respectively called the first quartile (Q1), the second quartile (Q2, also known as the median), and the third quartile (Q3). Specifically, the first quartile (Q1) represents the value at the 25% position after the data set is arranged in ascending order. The second quartile (Q2) represents the value at the 50% position after the data set is arranged in ascending order, that is, the median. The third quartile (Q3) represents the value at the 75% position after the data set is arranged in ascending order. Using the interquartile range, the range of outliers can be defined: Interquartile range:

[0025] Lower bound:

[0026] Upper bound:

[0027] Among them, Q1 represents the first quartile, Q3 represents the third quartile, and IQR represents the interquartile range.

[0028] In the original input data, any data point below the lower bound or above the upper bound is determined as an outlier.

[0029] Furthermore, the detected outlier data points are set to null values, and the null values and missing values are filled. Missing value filling can be divided into three categories: head missing values, tail missing values, and middle missing values. Among them, head missing values can be filled using backward fill (BFILL), tail missing values are filled using forward fill (FFILL), and middle missing values can be filled using linear interpolation. Linear interpolation is used to estimate the value of unknown points between known data points. For example: Assume that between two known data points, the value of the function changes linearly. Assume that two known data points are (x1, y1) and (x2, y2), where x1 < x2, and the function value y at an unknown point x between x1 and x2 needs to be estimated. Linear interpolation performs linear interpolation between y1 and y2 based on the relative position of x to x1 and x2 to obtain an estimated value of y. The calculation formula for linear interpolation is:

[0030] Among them, y represents the estimated value at data point x, and y1 and y2 respectively represent the estimated values corresponding to known data points x1 and x2.

[0031] S102: Train the initial electricity price prediction model based on the second feature set to obtain the target electricity price prediction model.

[0032] like Figure 2 As shown, in this embodiment, the initial electricity price prediction model is a CRFNet deep neural network built based on the Pytroch deep learning framework. The full name of CRFNet is Conv Recursive Full-Connection Net. The CRFNet deep neural network is constructed as a Model class. The CRFNet deep neural network consists of 4 sub-modules, including an autoregressive module (AR-Block), a continuous sampling module (C-Block), an interval sampling module (I-Block) and a convolutional recurrent module (CRN-Block).

[0033] Exemplarily, the second feature set is a feature set that has an important impact on electricity price prediction after screening and processing. The second feature set is input into the initial electricity price prediction model (CRFNet deep neural network), which continuously adjusts its own parameters by learning the complex relationship between these features and electricity prices to improve the accuracy of electricity price prediction. CRFNet consists of multiple submodules, each of which analyzes and processes the input features from different angles, and works together to capture various factors of electricity price changes.

[0034] During the training process, the initial electricity price prediction model uses the second feature set as input and the actual market clearing electricity price as the target output. By continuously adjusting the model parameters, the model's predicted output is made as close to the actual electricity price as possible. The Pytorch deep learning framework provides efficient computing and optimization algorithms that can automatically calculate the model's loss function (such as mean square error, etc.) and update the model parameters based on the gradient information of the loss function to minimize the prediction error.

[0035] S103: Obtaining an electricity price prediction result for a second time after the time point to be measured based on historical data for a first time before the time point to be measured and a target electricity price prediction model.

[0036] In this embodiment, the time point to be tested refers to the specific moment when electricity price prediction is required in the future, which is expressed as a specific date and time, such as January 1, 2025. The first time is a time period related to the time point to be tested, which is used to determine the scope of historical data, and can be in units of days, weeks, months, etc., such as two weeks before the time point to be tested. Parameters may include time length units, specific durations, etc. Historical data refers to electricity price-related data before the time point to be tested. It may include load data (such as electricity consumption in different time periods), environmental data (such as temperature, season, etc.), etc. Parameters may include specific values ​​of multiple features and corresponding timestamps. The second time is the time period corresponding to the predicted future electricity price, such as 45 days. The electricity price prediction result is an estimated value of the electricity price in the second time in the future.

[0037] In this embodiment, the target electricity price prediction model can be used to learn the relationship between various features in historical data and electricity prices, and the electricity price at the second time after the test time point is inferred by analyzing the historical data of the first time before the test time point.

[0038] Exemplarily, determine the length of the first time and the prediction range of the second time. Collect and organize historical data of the first time before the time point to be tested, including multi-dimensional data such as load and environment, and perform preprocessing. Input these data into the target electricity price prediction model, and each submodule extracts the characteristics of historical data from different angles. After calculation and comprehensive analysis of the target electricity price prediction model, output the electricity price prediction result of the second time after the time point to be tested, and provide decision-making basis for electricity market participants.

[0039] From the above, it can be concluded that, on the one hand, this embodiment deeply analyzes the correlation between the multi-dimensional market boundary condition characteristics and the market clearing electricity price, selects the characteristics with strong correlation and performs data processing. This step effectively integrates the multi-dimensional factors affecting the electricity price, lays a solid foundation for the subsequent electricity price prediction model training, reduces the interference of irrelevant information, and makes the prediction results more accurate.

[0040] On the other hand, this embodiment uses the processed feature set data to train the initial electricity price prediction model, which can fully capture the complex laws of market supply and demand changes, thereby generating a high-precision target electricity price prediction model. The target electricity price prediction model not only considers the continuity of the time series, but also incorporates the correlation between multi-dimensional features, significantly improving the comprehensiveness and accuracy of the prediction.

[0041] On the other hand, based on the historical data before the test time point and the trained target electricity price prediction model, the electricity price data within a longer time period can be reliably predicted, providing strong decision-making support for power dispatching agencies, and helping to achieve a more reasonable allocation of power resources and smooth operation of the market.

[0042] In one embodiment of the present disclosure, the multi-dimensional market boundary condition characteristics include a load characteristic set and / or an environmental characteristic set.

[0043] The first feature set is determined based on the correlation between the multi-dimensional market boundary condition characteristics and the market clearing electricity price, including: The correlation coefficients between the load characteristic set and / or the environmental characteristic set and the market clearing electricity price are calculated to obtain a correlation coefficient matrix.

[0044] A first feature set is selected from the load feature set and / or the environment feature set based on the correlation coefficient matrix.

[0045] In this embodiment, selecting a first feature set from a load feature set and / or an environment feature set based on a correlation coefficient matrix includes: If the correlation coefficient in the correlation coefficient matrix is ​​greater than a first threshold, the correlation coefficient is marked as a target correlation coefficient, and features in a load feature set and / or an environment feature set corresponding to the target correlation coefficient are counted as a first feature set.

[0046] In this embodiment, the correlation coefficient between the load characteristic set and / or the environmental characteristic set and the market clearing electricity price is calculated to obtain a correlation coefficient matrix, including: calculating the correlation coefficient between the load characteristic set and / or the environmental characteristic set and the market clearing electricity price based on the Pearson correlation coefficient, and obtaining a correlation coefficient matrix based on the correlation coefficient.

[0047] Exemplarily, after collecting the required multidimensional market boundary condition characteristics, the Pearson Correlation Coefficient is used to analyze the correlation between the multidimensional market boundary condition characteristics and the market clearing electricity price. The Pearson correlation coefficient is used to measure the degree of linear correlation between two variables. The value range of the Pearson correlation coefficient is between -1 and 1, where 1 indicates a perfect positive correlation, that is, there is a perfect linear relationship between the feature and the target; -1 indicates a perfect negative correlation, that is, there is a perfect linear relationship between the feature and the target, and when the feature value of a certain variable increases, the target value decreases. 0 means there is no linear correlation, that is, there is no linear relationship between the two variables. According to the calculation results of the Pearson correlation coefficient, it can be decided to select features with correlation coefficient values ​​in the range of (-0.4~-1 and 0.4~1). The calculation formula for the Pearson correlation coefficient is:

[0048] Where r represents the correlation coefficient, represents the i-th sample data of the X variable, represents the i-th sample data of the Y variable, n represents the number of sample data of the X variable or the Y variable, represents the mean of sample data X, Represents the mean of the sample data Y.

[0049] In this embodiment, the first threshold is used to measure the strength of the correlation between the feature and the market-clearing electricity price. The first threshold can be determined based on experience or by testing different values ​​and evaluating model performance. When the correlation coefficient exceeds this first threshold, it is considered that the corresponding feature has a strong enough correlation with the electricity price and is worthy of being selected into the feature set for subsequent electricity price prediction. The target correlation coefficient refers to those correlation coefficients in the correlation coefficient matrix that are greater than the first threshold. The features corresponding to these coefficients are considered to have an important impact on the market-clearing electricity price. Marking these coefficients as target correlation coefficients is to facilitate the subsequent screening of the corresponding key features.

[0050] This embodiment can effectively identify the key factors affecting electricity prices by analyzing the correlation between the characteristics of multi-dimensional market boundary conditions and the market clearing electricity price. The Pearson correlation coefficient can be used to quantify the strength of the linear relationship between the load feature set and the environmental feature set and the market clearing electricity price. The construction of the correlation coefficient matrix helps to screen out the features that are highly correlated with electricity price changes, thereby providing decision support for market participants, optimizing resource allocation, enhancing market transparency, and promoting the stability and efficiency of the electricity market.

[0051] In one embodiment of the present disclosure, the target electricity price prediction model includes: a continuous sampling module, an interval sampling module, a circular convolution module and an autoregressive module.

[0052] In this embodiment, the electricity price prediction result at the second time after the time point to be tested is obtained based on the historical data at the first time before the time point to be tested and the target electricity price prediction model, including: The historical data of the first time before the time point to be tested is input into the continuous sampling layer to obtain the multi-dimensional feature deep temporal relationship and continuity relationship of the historical data.

[0053] The historical data of the first time before the time point to be tested and the market-clearing electricity price corresponding to the historical data are input into the interval sampling layer to obtain the correlation characteristics between the historical data and the market-clearing electricity price.

[0054] The historical data of the first time before the time point to be tested is input into the convolutional loop layer to obtain the deep sequence features of the historical data, and the deep sequence features are input into the long short-term memory network to extract the long short-term memory correlation data of the deep sequence features.

[0055] The historical data before the time point to be tested is input into the autoregressive layer to obtain the time linear characteristics of the historical data itself.

[0056] The deep temporal relationship and continuity relationship of the multi-dimensional characteristics of historical data, the correlation characteristics between historical data and market clearing electricity prices, long-term and short-term memory correlation data and the time linear characteristics of historical data themselves are aggregated to obtain the electricity price forecast result for the second time after the test time point.

[0057] In this embodiment, the historical data of the first time before the time point to be measured is input into the continuous sampling layer to obtain the multi-dimensional feature deep temporal relationship and continuity relationship of the historical data, including: The historical data of the first time before the time point to be tested is input into the continuous sampling layer, and the historical data is divided based on the T / C dimension and the C dimension. The C dimension represents the sequence division length, and the T / C dimension represents the number of sequences after division.

[0058] The divided historical data are spliced ​​to obtain C (T / C) feature matrix.

[0059] C-based The feature matrix of (T / C) obtains the multi-dimensional features, deep temporal relationships and continuity relationships of historical data.

[0060] In this embodiment, the continuous sampling layer is composed of two upper and lower feature sampling layers, wherein the upper layer includes two continuous feature extraction fully connected layers (Continuous Full Connection Layer, CFC Layer). CFC Layer_1 includes parameters input_dim and hidden_size, and CFC Layer_2 includes parameters hidden_size and output_size. The lower layer is a channel feature extraction fully connected layer (Channel Full Connection Layer, Channel FC). Channel FC_1 includes parameters channel_input and channel_output. A Channel FC_2 is set at the end of the continuous sampling layer, and Channel FC_2 includes parameters output_size and output_dim. The input feature data dimension of the continuous sampling layer is [B, T, N]. First, the feature dimension is converted into [B, T / C, C, N] through dimensionality transformation, and the B dimension and N dimension are merged to obtain [B N, C, T / C], will [B N, C, T / C] is input into the lower layer to extract channel features, and [B N, T / C, C] input to the upper layer to extract continuous features, and finally add the two features and get the final output through Channel FC_2. The original input features (i.e., multi-dimensional market boundary condition features) can be extracted through the continuous sampling layer to extract the deep temporal relationship and continuous relationship of the multi-dimensional features in the time dimension, and introduce the seasonality and trend characteristics of the multi-dimensional feature sequence itself for electricity price prediction (i.e., the deep temporal relationship and continuous relationship of the multi-dimensional features of historical data).

[0061] In this embodiment, the interval sampling layer is composed of two upper and lower feature sampling layers, wherein the upper layer includes two interval feature extraction fully connected layers (Interval Full Connection Layer, IFC Layer). Among them, IFC Layer_1 includes parameters input_dim and hidden_size, and IFC Layer_2 includes parameters hidden_size and output_size. The lower layer is a channel feature extraction fully connected layer (Channel Full Connection Layer, Channel FC) Channel FC_3, and Channel FC_3 includes parameters channel_input and channel_output. A layer of Channel FC_4 is set at the end of the interval sampling layer, and Channel FC_4 includes parameters output_size and output_dim. The input feature data dimension of the interval sampling layer is [B, T, N]. First, the feature dimension is converted into [B, C, T / C, N] through dimensionality transformation, and then the B dimension and N dimension are merged to obtain [B N, C, T / C], will [B N, C, T / C] is input into the lower layer to extract channel features, and [B N, T / C, C] is input into the upper layer to extract interval features, and finally the features of the two outputs are added and the final output is obtained through Channel FC_4. The original input features are passed through the interval sampling layer, which expands the receptive field of feature extraction compared to continuous sampling, splits a single sequence into multiple equal-length sequences for splicing, transforms the dimension of the original feature sequence, and extracts features horizontally to explore the correlation between deep boundary condition features, weather features, and market clearing electricity prices (that is, the correlation features between historical data and market clearing electricity prices).

[0062] In this embodiment, the convolutional loop layer includes a one-dimensional convolutional layer Conv1D, a long short-term memory network (Long - Short Term Memory, LSTM), a channel fully connected layer Channel FC, and a sequence fully connected layer Sequence FC. Conv1D includes parameters in_channels, out_channels, kernel_size, and padding; LSTM includes parameters input_size and hidden_siz; the channel fully connected layer Channel FC includes parameters in_features and out_features; and the sequence fully connected layer Sequence FC includes parameters in_features and out_features. The original feature input is fed into the convolutional loop layer, and the original feature is downsampled through a one-dimensional convolution module to obtain a deep sequence feature, which is then input into the LSTM module to extract the long-term memory correlation and short-term memory correlation of the deep sequence feature, and output a deep time series feature with long-term and short-term correlation (i.e., the long-term and short-term memory correlation data of the deep sequence feature).

[0063] In this embodiment, the autoregressive layer includes two autoregressive linear layers AR Linear, and AR Linear includes parameters seq_len and pred_len. The autoregressive layer can extract the time linear relationship of the original feature sequence itself (that is, the time linear feature of the historical data itself), and introduce linear features into the nonlinear deep neural network.

[0064] For example, a power trading company needs to predict the electricity price in the next month (the second time after the test time point) in order to formulate a trading strategy. The company collects historical data from the past three months (the first time before the test time point), including power load data in various places (load conditions in different cities and different time periods), environmental data (temperature, weather conditions, etc.) and the corresponding market clearing electricity price.

[0065] Input these data into the target electricity price prediction model. In the continuous sampling layer, the data is divided according to the T / C and C dimensions, such as by day and week, to extract the seasonal and trend characteristics of the load data in the time dimension, such as the peak electricity consumption trend in summer. The interval sampling layer mines the association between hot weather (environmental characteristics), peak electricity consumption (load characteristics) and high electricity prices. The convolutional recurrent layer uses convolution and LSTM to extract long-term and short-term memory correlations such as the periodic changes in industrial electricity consumption. The autoregressive layer obtains the linear relationship between daily electricity load and time. Aggregating these features predicts the electricity price for the next month, and the company formulates a reasonable execution strategy based on this.

[0066] This embodiment uses modules such as continuous sampling, interval sampling, convolution loop and autoregression to comprehensively and deeply mine the multidimensional characteristics, temporal relationships, correlations and long-term and short-term memory correlations of historical data, so as to accurately predict future electricity prices. This not only provides a scientific decision-making basis for power trading companies and helps them formulate reasonable trading strategies, but also significantly improves the accuracy and reliability of electricity price forecasts. In addition, the target electricity price forecasting model has strong adaptability and flexibility, and can cope with the electricity price forecasting needs in different markets and environments, providing strong support for the stable development of the power market.

[0067] In one embodiment of the present disclosure, a medium- and long-term electricity price forecasting method further includes: determining the power supply status based on the medium- and long-term electricity price forecasting results.

[0068] Determine resource allocation and management strategies based on power supply conditions.

[0069] In this embodiment, the medium- and long-term electricity price forecast results reflect the price trend of the electricity market in the future. For example, an increase in electricity prices may mean that electricity supply is tight, because the shortage of supply will drive up prices; while a decrease in electricity prices may indicate that electricity supply is relatively sufficient, and supply exceeds demand, resulting in price drops. By analyzing the level and trend of electricity price forecasts, it can be inferred whether the electricity supply is sufficient, tight, or balanced. After determining the power supply status, resource allocation and management strategies can be adjusted according to different supply conditions. For example, when the supply is tight, it is necessary to reasonably allocate power generation resources and optimize grid operation to ensure power consumption for key users; when the supply is sufficient, it is possible to consider optimizing resource utilization efficiency and performing equipment maintenance.

[0070] For example, a threshold of electricity price can be set. When the predicted medium- and long-term electricity price is higher than the threshold, it is judged that the electricity supply is tight; when the electricity price is lower than the threshold, it is judged that the electricity supply is sufficient; and between the two, it is a balance between supply and demand. At the same time, factors such as the rate of change of electricity price can be combined to make a more accurate judgment. For example, if it is predicted that the electricity price will continue to rise in the next 30 days and exceed the set tight threshold, it indicates that the electricity supply is in a tight trend.

[0071] For example, when supply is tight: for power generation companies, the power generation capacity of high-efficiency units can be increased to prioritize the supply of power generation fuel; for power grid companies, the monitoring and dispatching of the power grid can be strengthened, network losses can be reduced by adjusting the power flow distribution, priority can be given to power supply to important users such as hospitals and transportation hubs, and orderly power consumption plans can be implemented when necessary to limit the power consumption of some non-critical users.

[0072] When the supply is sufficient: power generation companies can arrange maintenance and overhaul of some units to improve equipment reliability; power grid companies can upgrade and transform the power grid, optimize the power grid topology, and improve the transmission capacity and stability of the power grid. At the same time, some preferential electricity price strategies can be introduced to promote industrial production.

[0073] When supply and demand are balanced: maintain existing power generation and supply plans, continuously monitor market dynamics, optimize resource utilization efficiency, and improve the overall economy and environmental performance of the power system.

[0074] This embodiment can provide early insight into the supply and demand situation of the electricity market by accurately predicting electricity price trends, and provide a scientific basis for power supply management. When the power supply is tight, it can quickly adjust resource allocation to ensure power supply to key users and maintain social stability; when the supply is sufficient, it can optimize resource utilization efficiency, perform equipment maintenance and upgrades, and improve the stability of the power system. At the same time, reasonable strategies are formulated based on the supply and demand balance to improve overall economy and environmental protection. This embodiment not only enhances the flexibility and resilience of the power system, but also helps to achieve sustainable development of the power market.

[0075] Corresponding to a medium- and long-term electricity price forecasting method in the above embodiment, Figure 3 This is a structural block diagram of a medium- and long-term electricity price prediction device provided by an embodiment of the present disclosure. For ease of explanation, only the parts related to the embodiment of the present disclosure are shown. Figure 3 The medium- and long-term electricity price prediction device 20 includes: a feature extraction module 21, a model training module 22 and an electricity price prediction module 23.

[0076] The feature extraction module 21 is used to determine a first feature set based on the correlation between the multi-dimensional market boundary condition characteristics and the market clearing electricity price, and to perform data processing on the first feature set to obtain a second feature set.

[0077] The model training module 22 is used to train the initial electricity price prediction model based on the second feature set to obtain a target electricity price prediction model.

[0078] The electricity price prediction module 23 is used to obtain the electricity price prediction result at a second time after the time point to be tested based on the historical data at a first time before the time point to be tested and the target electricity price prediction model.

[0079] In one embodiment of the present disclosure, the multi-dimensional market boundary condition characteristics include a load characteristic set and / or an environmental characteristic set.

[0080] The feature extraction module 21 is specifically used to calculate the correlation coefficient between the load feature set and / or the environmental feature set and the market clearing electricity price to obtain a correlation coefficient matrix.

[0081] A first feature set is selected from the load feature set and / or the environment feature set based on the correlation coefficient matrix.

[0082] In one embodiment of the present disclosure, the feature extraction module 21 is specifically used to mark the correlation coefficient as a target correlation coefficient if the correlation coefficient in the correlation coefficient matrix is ​​greater than a first threshold, and to count the features in the load feature set and / or environmental feature set corresponding to the target correlation coefficient as a first feature set.

[0083] In one embodiment of the present disclosure, the target electricity price prediction model includes: a continuous sampling module, an interval sampling module, a circular convolution module and an autoregressive module.

[0084] In one embodiment of the present disclosure, the electricity price prediction module 23 is specifically used to input the historical data of the first time before the time point to be measured into the continuous sampling layer to obtain the multi-dimensional feature deep temporal relationship and continuity relationship of the historical data.

[0085] The historical data of the first time before the time point to be tested and the market-clearing electricity price corresponding to the historical data are input into the interval sampling layer to obtain the correlation characteristics between the historical data and the market-clearing electricity price.

[0086] The historical data of the first time before the time point to be tested is input into the convolutional loop layer to obtain the deep sequence features of the historical data, and the deep sequence features are input into the long short-term memory network to extract the long short-term memory correlation data of the deep sequence features.

[0087] The historical data before the time point to be tested is input into the autoregressive layer to obtain the time linear characteristics of the historical data itself.

[0088] The deep temporal relationship and continuity relationship of the multi-dimensional characteristics of historical data, the correlation characteristics between historical data and market clearing electricity prices, long-term and short-term memory correlation data and the time linear characteristics of historical data themselves are aggregated to obtain the electricity price forecast result for the second time after the test time point.

[0089] In one embodiment of the present disclosure, the electricity price prediction module 23 is further used to input the historical data of the first time before the time point to be measured into the continuous sampling layer, and divide the historical data based on the T / C dimension and the C dimension, the C dimension represents the sequence division length, and the T / C dimension represents the number of sequences after division.

[0090] The divided historical data are spliced ​​to obtain C (T / C) feature matrix.

[0091] C-based The feature matrix of (T / C) obtains the multi-dimensional features, deep temporal relationships and continuity relationships of historical data.

[0092] In one embodiment of the present disclosure, a medium- and long-term electricity price prediction device 20 further includes: The resource management module is used to determine the power supply status based on the medium- and long-term electricity price forecast results.

[0093] Determine resource allocation and management strategies based on power supply conditions.

[0094] See also Figure 4 , Figure 4 A schematic block diagram of an electronic device provided by an embodiment of the present disclosure. Figure 4 The electronic device 300 in the embodiment shown may include: one or more processors 301, one or more input devices 302, one or more output devices 303 and one or more memories 304. The processors 301, input devices 302, output devices 303 and memories 304 communicate with each other via a communication bus 305. The memory 304 is used to store computer programs, which include program instructions. The processor 301 is used to execute the program instructions stored in the memory 304. The processor 301 is configured to call the program instructions to execute the functions of the modules in the above-mentioned device embodiments, such as Figure 3 The functions of modules 21 to 23 are shown.

[0095] It should be understood that in the embodiment of the present disclosure, the processor 301 may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0096] The input device 302 may include a touch panel, a fingerprint collection sensor (for collecting the user's fingerprint information and fingerprint direction information), a microphone, etc., and the output device 303 may include a display (LCD, etc.), a speaker, etc.

[0097] The memory 304 may include a read-only memory and a random access memory, and provide instructions and data to the processor 301. A portion of the memory 304 may also include a non-volatile random access memory. For example, the memory 304 may also store information about the device type.

[0098] In a specific implementation, the processor 301, input device 302, and output device 303 described in the embodiments of the present disclosure can execute the implementation methods described in the first and second embodiments of a medium- and long-term electricity price forecasting method provided in the embodiments of the present disclosure, and can also execute the implementation methods of the electronic device 300 described in the embodiments of the present disclosure, which will not be repeated here.

[0099] In another embodiment of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by the processor, all or part of the processes in the above-mentioned embodiment method are implemented, and the computer program can also be completed by instructing the relevant hardware through the computer program. The computer program can be stored in a computer-readable storage medium, and when the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electric carrier signal, telecommunication signal and software distribution medium, etc.

[0100] The computer-readable storage medium may be an internal storage unit of the electronic device of any of the aforementioned embodiments, such as a hard disk or memory of the electronic device. The computer-readable storage medium may also be an external storage device of the electronic device, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (SecureDigital, SD) card, a flash card (Flash Card), etc. equipped on the electronic device. Further, the computer-readable storage medium may also include both an internal storage unit of the electronic device and an external storage device. The computer-readable storage medium is used to store computer programs and other programs and data required by the electronic device. The computer-readable storage medium may also be used to temporarily store data that has been output or is to be output.

[0101] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this disclosure.

[0102] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the electronic devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0103] In the several embodiments provided in the present application, it should be understood that the disclosed electronic devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces or units, or it can be an electrical, mechanical or other form of connection.

[0104] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present disclosure.

[0105] In addition, each functional unit in each embodiment of the present disclosure may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0106] The above are only specific embodiments of the present disclosure, but the protection scope of the present disclosure is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present disclosure, and these modifications or replacements should be included in the protection scope of the present disclosure. Therefore, the protection scope of the present disclosure should be based on the protection scope of the claims.

Claims

1. A medium- and long-term electricity price forecasting method, characterized in that: include: Determine a first feature set based on the correlation between the multi-dimensional market boundary condition characteristics and the market clearing electricity price, and perform data processing on the first feature set to obtain a second feature set; Training the initial electricity price prediction model based on the second feature set to obtain a target electricity price prediction model; The electricity price prediction result at a second time after the time point to be measured is obtained based on the historical data of the first time before the time point to be measured and the target electricity price prediction model.

2. A medium- and long-term electricity price forecasting method as claimed in claim 1, characterized in that: The multi-dimensional market boundary condition characteristics include a load characteristic set and / or an environmental characteristic set; The first feature set is determined based on the correlation between the multi-dimensional market boundary condition characteristics and the market clearing electricity price, including: Calculating the correlation coefficient between the load characteristic set and / or the environmental characteristic set and the market clearing electricity price to obtain a correlation coefficient matrix; The first feature set is selected from the load feature set and / or the environment feature set based on the correlation coefficient matrix.

3. A medium- and long-term electricity price forecasting method as claimed in claim 2, characterized in that: The selecting the first feature set from the load feature set and / or the environment feature set based on the correlation coefficient matrix comprises: If the correlation coefficient in the correlation coefficient matrix is ​​greater than a first threshold, the correlation coefficient is marked as a target correlation coefficient, and features in the load feature set and / or the environment feature set corresponding to the target correlation coefficient are counted as the first feature set.

4. A medium- and long-term electricity price forecasting method as claimed in claim 1, characterized in that: The target electricity price prediction model includes: a continuous sampling module, an interval sampling module, a circular convolution module and an autoregressive module.

5. A medium- and long-term electricity price forecasting method as claimed in claim 4, characterized in that: The step of obtaining the electricity price prediction result at a second time after the time point to be measured based on the historical data at a first time before the time point to be measured and the target electricity price prediction model includes: Inputting the historical data of the first time before the time point to be measured into the continuous sampling layer to obtain the multi-dimensional feature deep temporal relationship and continuity relationship of the historical data; Inputting historical data of the first time before the time point to be tested and the market-clearing electricity price corresponding to the historical data into the interval sampling layer to obtain the correlation characteristics between the historical data and the market-clearing electricity price; Inputting the historical data of the first time before the time point to be measured into the convolutional loop layer to obtain the deep sequence features of the historical data, inputting the deep sequence features into the long short-term memory network, and extracting the long short-term memory correlation data of the deep sequence features; Inputting the historical data of the first time before the time point to be measured into the autoregressive layer to obtain the time linear characteristics of the historical data itself; The deep temporal relationship and continuity relationship of the multi-dimensional characteristics of the historical data, the correlation characteristics between the historical data and the market clearing electricity price, the long-term and short-term memory correlation data and the time linear characteristics of the historical data themselves are aggregated to obtain the electricity price prediction result for the second time after the time point to be tested.

6. A method for predicting medium- and long-term electricity prices as claimed in claim 5, characterized in that: The step of inputting the historical data of the first time before the time point to be measured into the continuous sampling layer to obtain the multi-dimensional feature deep temporal relationship and continuity relationship of the historical data includes: Inputting the historical data of the first time before the time point to be tested into the continuous sampling layer, dividing the historical data based on the T / C dimension and the C dimension, the C dimension represents the sequence division length, and the T / C dimension represents the number of sequences after division; The divided historical data are spliced ​​to obtain C (T / C) feature matrix; Based on the C The characteristic matrix of (T / C) obtains the multi-dimensional characteristic deep temporal relationship and continuity relationship of the historical data.

7. A method for predicting medium- and long-term electricity prices as claimed in claim 1, characterized in that: Also includes: Determine the power supply status based on the medium- and long-term power price forecast results; A resource allocation and management strategy is determined based on the power supply status.

8. A medium- and long-term electricity price forecasting device, characterized in that: include: A feature extraction module, used to determine a first feature set based on the correlation between the multi-dimensional market boundary condition features and the market clearing electricity price, and perform data processing on the first feature set to obtain a second feature set; A model training module, used for training the initial electricity price prediction model based on the second feature set to obtain a target electricity price prediction model; The electricity price prediction module is used to obtain the electricity price prediction result of the second time after the time point to be tested based on the historical data of the first time before the time point to be tested and the target electricity price prediction model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images