Electric power short-term load hybrid prediction method, system, equipment and medium

Abstract

The invention provides an electric power short-term load hybrid prediction method, system and device, and a medium. The method comprises the following steps: collecting related electric power load and meteorological data as a first data set; obtaining an optimal combination of a modal number K and a penalty parameter alpha of the signal decomposition algorithm by adopting an optimization algorithm; decomposing the first data set into a plurality of signal components based on the optimal combination; calculating the coefficients of the signal components by using the related performance indexes, and selecting the most related signal component to construct an input matrix; and determining a zero-crossing rate separation point according to feature information in the input matrix, realizing separation of high and low frequencies of the signal components according to the zero-crossing rate separation point, and inputting the high-frequency signal components and the low-frequency signal components into a time sequence prediction model to obtain prediction results of each high-frequency signal component and each low-frequency signal component, and obtaining a power short-term load hybrid prediction result. According to the technical scheme, the prediction accuracy can be effectively improved, and safe and stable operation of the power system is facilitated.

Description

Technical Field

[0001] This document relates to the technical field of short-term power load forecasting, and in particular to a method, system, equipment and medium for hybrid short-term power load forecasting. Background Art

[0002] Electricity demand is affected by many factors such as weather, electricity prices and accidents, resulting in random fluctuations in electricity load. Therefore, in order to reduce random fluctuations in electricity load, the power load of the power grid can be predicted, and the prediction results of the power load can be used to adjust the balance between the power supply plan and electricity demand of the power grid. According to the length of the prediction time, power load forecasting is divided into long-term load forecasting (LTLF), medium-term load forecasting (MTLF) and short-term load forecasting (STLF). Accurate short-term load forecasting can not only enable the normal operation of the system and reduce energy waste, but also improve power supply reliability.

[0003] In the prior art, the Informer model is usually used as the time series prediction model. Based on the Transformer model, the Informer model effectively reduces the time and space complexity, memory usage and decoder decoding time of the Transformer model. In addition, it also improves the prediction accuracy of the time series and has a stronger ability to capture the long-range correlation coupling between the input and output data. At the same time, a variety of combined prediction methods are widely used in short-term load forecasting. The decomposition of time series usually uses the Empirical Mode Decomposition (EMD) algorithm, the Ensemble Empirical Mode Decomposition (EEMD) algorithm and the Variational Mode Decomposition (VMD) algorithm to decompose the time series data, but the above time series decomposition algorithms have the following disadvantages:

[0004] Empirical Mode Decomposition (EMD) can decompose the original load sequence. Ensemble Empirical Mode Decomposition (EEMD) can effectively alleviate the modal aliasing phenomenon of EMD by adding Gaussian white noise before decomposition. However, the robustness of EMD and EEMD to measurement noise needs to be strengthened.

[0005] Variational Mode Decomposition (VMD) decomposes data into modal functions with different characteristics, effectively avoiding modal aliasing. However, the effectiveness of load sequence decomposition is affected by the number of modal components and penalty parameters. In the field of power load forecasting, there is a lack of scientific evaluation criteria for determining these parameters, which are often determined based on experience and subjectivity, affecting the effect and overall accuracy of load sequence decomposition. Summary of the invention

[0006] The present invention provides a method, system, device and medium for hybrid forecasting of short-term power load, aiming to solve the above-mentioned problem.

[0007] The embodiment of the present invention provides a hybrid forecasting method for short-term power load, including:

[0008] S1. Collect relevant power load and meteorological data as a first data set;

[0009] S2. Using a preset optimization algorithm to obtain an optimal combination of the modal number K and the penalty parameter α of the signal decomposition algorithm;

[0010] S3, decomposing the first data set into a plurality of signal components using a signal decomposition algorithm based on an optimal combination, wherein the signal components have different frequencies and different amplitudes;

[0011] S4, using the correlation performance index to calculate the maximum mutual information coefficient of each signal component, and selecting the most relevant features of a preset number of signal components according to the maximum mutual information coefficient of each signal component to construct an input matrix;

[0012] S5, using an adaptive algorithm to determine a zero-crossing rate separation point according to the characteristic information in the input matrix, and separating the high and low frequencies of the signal components according to the zero-crossing rate separation point to obtain a high-frequency signal component and a low-frequency signal component;

[0013] S6. Input the high-frequency signal component and the low-frequency signal component into a time series prediction model to obtain a prediction result of each high-frequency signal component and low-frequency signal component, and reconstruct the prediction result of each high-frequency signal component and low-frequency signal component to obtain a prediction result of the short-term power load mixture.

[0014] The embodiment of the present invention provides a hybrid short-term power load forecasting system, including:

[0015] A data collection module, used for collecting relevant power load and meteorological data as a first data set;

[0016] An optimal parameter acquisition module is used to obtain an optimal combination of the mode number K and the penalty parameter α of the signal decomposition algorithm using a preset optimization algorithm;

[0017] A signal component acquisition module is configured to decompose the first data set into a plurality of signal components by using a signal decomposition algorithm based on an optimal combination, wherein the signal components have different frequencies and different amplitudes;

[0018] Obtaining an input matrix module, calculating the maximum mutual information coefficient of each signal component using the relevant performance index, and selecting the most relevant features of a preset number of signal components according to the maximum mutual information coefficient of each signal component to construct an input matrix;

[0019] A high- and low-frequency signal separation module is used to determine a zero-crossing rate separation point using an adaptive algorithm according to the characteristic information in the input matrix, and to separate the high and low frequencies of the signal components according to the zero-crossing rate separation point to obtain a high-frequency signal component and a low-frequency signal component;

[0020] The result prediction module is used to input the high-frequency signal component and the low-frequency signal component into the time series prediction model to obtain the prediction result of each high-frequency signal component and the low-frequency signal component, and reconstruct the prediction result of each high-frequency signal component and the low-frequency signal component to obtain the prediction result of the short-term power load mixture.

[0021] An embodiment of the present invention provides an electronic device, including:

[0022] processor; and,

[0023] A memory is arranged to store computer executable instructions, which, when executed, cause the processor to perform the steps of the above-mentioned power short-term load hybrid forecasting method.

[0024] An embodiment of the present invention provides a storage medium for storing computer executable instructions, which, when executed, implement the steps of the above-mentioned power short-term load hybrid forecasting method.

[0025] According to an embodiment of the present invention, a storage medium is provided for storing computer executable instructions, wherein the computer executable instructions, when executed, implement the steps described in the above method embodiment.

[0026] The embodiment of the present invention obtains the optimal combination of the modal number K and the penalty parameter α of the signal decomposition algorithm through a preset optimization algorithm, decomposes the data in the first data set including the relevant power load and meteorological data based on the signal decomposition algorithm of the optimal combination, thereby obtaining the signal component with the best effect, and performs feature screening on the signal component to obtain the input matrix, determines the high-frequency and low-frequency signal dividing points of the signal component with the zero-crossing rate as an indicator to obtain the high-frequency signal and the low-frequency signal, predicts the high-frequency signal component and the low-frequency signal component respectively, and reconstructs all the prediction results to obtain the hybrid prediction result of the short-term power load. The hybrid prediction method of the short-term power load provided by the embodiment of the present invention can effectively improve the prediction accuracy, which is beneficial to the safe and stable operation of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, 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 recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0028] Figure 1 A flow chart of a hybrid power short-term load forecasting method according to an embodiment of the present invention;

[0029] Figure 2 A flowchart of optimizing VMD parameters for FSA according to an embodiment of the present invention;

[0030] Figure 3 is an overall flow chart of the time series prediction model of an embodiment of the present invention;

[0031] Figure 4 Schematic diagram of a hybrid power short-term load forecasting system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this document.

[0033] Method Embodiment

[0034] According to an embodiment of the present invention, a method for hybrid short-term power load forecasting is provided. Figure 1 The flowchart of the hybrid forecasting method for short-term power load according to the embodiment of the present invention is as follows: Figure 1 As shown, the hybrid short-term power load forecasting method of the embodiment of the present invention specifically includes:

[0035] S1. Collect relevant power load and meteorological data as a first data set;

[0036] S2. Using a preset optimization algorithm to obtain an optimal combination of the modal number K and the penalty parameter α of the signal decomposition algorithm;

[0037] According to the embodiment of the present invention, this patent proposes FSA to optimize the optimal combination of VMD parameters. FSA combines the advantages of Firefly Algorithm (FA) and Sparrow Search Algorithm (SSA). Compared with SSA, FSA can effectively improve the global search capability and avoid falling into the local optimal solution. The specific steps of FSA to optimize VMD parameters are as follows: Figure 2 shown.

[0038] The specific process of FA and SSA forming FSA and optimizing VMD is as follows:

[0039] S21, randomly generate one or more parameter combinations [K, α] of modal number K and penalty parameter α as the initial positions of the discoverer and the follower;

[0040] S22, performing VMD decomposition on the initial positions of the finder and the follower, and calculating the fitness values ​​of the finder and the follower according to the decomposition results;

[0041] S23, updating the positions of the discoverer and the follower based on the alarm so that the discoverer and the follower escape from the dangerous area, and updating the positions of the sparrows based on the fitness values ​​of the discoverer and the follower obtained in S22;

[0042] S24, interfering with the sparrow position by the firefly algorithm to obtain a second sparrow position, if the second sparrow position is better than the sparrow position obtained in S23, then using the second sparrow position as a better sparrow position, otherwise, using the sparrow position obtained in S23 as the better sparrow position;

[0043] S25, using a better sparrow position to perform VMD decomposition on the updated finder and follower positions, and calculating the fitness values ​​of the finder and follower according to each decomposition result;

[0044] S26, S26, randomly select alarms and iteratively update the positions of the discoverers and followers, update the sparrow positions based on the fitness values ​​of the discoverers and followers obtained in S25, and determine whether the fitness values ​​of the discoverers and followers obtained in S25 meet the preset stop conditions. If so, obtain the minimum fitness value among the current fitness values ​​of the discoverers and followers, and obtain the modal number K and penalty parameter α corresponding to the minimum fitness value as the optimal combination [K, α], otherwise, execute step S23.

[0045] In the feature extraction part, VMD is used to decompose the original power load data. During the decomposition process, FSA is used to optimize the modal number K and penalty parameter α in VMD. The original power load data is decomposed into multiple subsequences through FSA-VMD, and then MIC is used to calculate the maximum mutual information coefficient of each subsequence to construct the input matrix of the model.

[0046] MIC is used to evaluate the correlation between two variables. Assume that X and Y are two different variables in the data set, where X = {x1, ..., x n}, Y={y1,....,y n}, n is the sample size. The maximum mutual information coefficient between X and Y is defined as:

[0047]

[0048] Where p(x,y) is the joint probability density between X and Y; p(x) and p(y) represent the marginal probability densities of X and Y respectively.

[0049] Draw a grid on the scatter plot of data consisting of variables X and Y, and calculate the size of the mutual information between grids. Use different grid division criteria to select the maximum value of mutual information M(x,y), which is calculated as follows:

[0050]

[0051] Where a and b are the number of grids divided in the X and Y directions respectively, and B is the maximum value of the grid.

[0052] The sparrow search algorithm is an intelligent optimization algorithm based on the predation behavior of sparrows. It optimizes the iterative update of the sparrows' positions by dividing them into discoverers, joiners, and guards. A discoverer is a sparrow that finds food, a joiner is a sparrow that follows the discoverer, and a guard is a sparrow that issues an early warning when it finds a predator. The sparrow population can be simulated by a matrix of n rows and d columns, where n represents the number of sparrows and d is the dimension of the problem to be optimized.

[0053] The expression for updating the position of the discoverer sparrow is:

[0054]

[0055] Where: t is the current iteration number; X i,j is the position information of the i-th sparrow in the j-th dimension; where i∈[1,c],j∈[1,d], the constant c is the population size; the constant d is the dimension to be optimized; α s is a random number that follows an average distribution of 0 to 1; Q is a random number that follows a Gaussian distribution; q max is the maximum number of iterations, which is a constant; L is a 1×d matrix, and all elements in the matrix are 1. R2 is a random number that follows an average distribution, representing the warning value; S T It is a random number with an average distribution of 0.5 to 1, representing a safety value. T It means that there are no natural enemies around when the sparrow group is foraging, and the discoverer will expand the search range and speed up the search; R2≥S T It means that the sparrow group has found a natural enemy. Both the discoverer and the joiner need to stay away from the natural enemy's activity range to forage. According to the joiner's behavior, update the joiner's position:

[0056]

[0057] A + =A T (AA T ) -1 ;

[0058] Where: X p , X worst are the best and worst positions occupied by sparrows during the iteration process; A is a 1×d matrix, and the matrix elements are assigned with average probabilities of 1 and -1. According to the reaction law of the sparrow group to the natural enemy, update the position where the sparrow group is aware of the danger:

[0059]

[0060] Where: β is a random number that conforms to the Gaussian distribution; U is a random number that obeys the average distribution of -1 to 1; it is used to control the step length and the direction of the sparrow's movement; f i is the fitness value of the current individual; f w is the global minimum fitness value; f g is the global maximum fitness value; U sets a very small constant to prevent the denominator from being 0.

[0061] ​The basic idea of ​​the firefly algorithm is that fireflies will move towards fireflies brighter than themselves, and will also be attracted by other fireflies. In this way, fireflies will gradually gather near a better solution, thereby achieving the purpose of optimization. The embodiment of the present invention combines the advantages of the firefly algorithm (FA) and the sparrow search algorithm (SSA).

[0062] S3, decomposing the first data set into a plurality of signal components using a signal decomposition algorithm based on an optimal combination, wherein the signal components have different frequencies and different amplitudes;

[0063] According to an embodiment of the present invention, the original power load data, i.e., the first data set, is decomposed using a VMD signal decomposition algorithm, and the original signal is decomposed into a number of intrinsic modal functions (IMFs), each of which has different frequency and amplitude characteristics. It can be seen from the decomposition steps of VMD that before decomposing the signal, it is necessary to set a suitable modal number K and penalty parameter α. Different combinations of K and α will have different effects on the decomposition results, and the optimal combination needs to be determined in combination with specific scenarios.

[0064] S4, using the correlation performance index to calculate the maximum mutual information coefficient of each signal component, and selecting the most relevant features of a preset number of signal components according to the maximum mutual information coefficient of each signal component to construct an input matrix;

[0065] S5, using an adaptive algorithm to determine a zero-crossing rate separation point according to the characteristic information in the input matrix, and separating the high and low frequencies of the signal components according to the zero-crossing rate separation point to obtain a high-frequency signal component and a low-frequency signal component;

[0066] After feature screening of the original load sequence, several IMF input matrices containing different impact features are obtained. These IMF input matrices have different time series information. In the field of electromagnetics, the zero-crossing rate is often used as the main feature to classify signals. In this regard, the zero-crossing rate is used as an evaluation index. High-frequency IMFs have a higher zero-crossing rate due to their faster rate of change, and low-frequency IMFs have a lower zero-crossing rate due to their slower rate of change. By calculating the zero-crossing rate of each IMF, sorting them from small to large, and using the separation point as the dividing point between high-frequency IMFs and low-frequency IMFs, the prediction error is calculated. By finding the optimal dividing point between high-frequency IMFs and low-frequency IMFs, the zero-crossing rate IMFs below the dividing point are regarded as low-frequency IMFs, and the zero-crossing rate IMFs above the dividing point are regarded as high-frequency IMFs.

[0067] Low-frequency signal components are usually slow and irregular, with different cycle lengths. To solve this problem, this patent uses an LSTM model for prediction. The LSTM prediction model is very suitable for processing irregular time series data and adapting to changing trends in low-frequency signals. Its memory unit structure allows it to store previous information and recall it in subsequent calculations, enabling it to capture historical information and effectively process trend items of low-frequency signals. In addition, the gating structure of the model helps to avoid the problem of long-term dependence on time series data, making it particularly suitable for predicting low-frequency time series components that are smooth, low-complexity, and have obvious periodicity in signal components.

[0068] The period of high-frequency signal components is very short, and their amplitude and frequency often change with changes in external factors or events, which is short-lived. The spectral entropy energy distribution tends to have concentrated high-frequency signal components. In this regard, this patent uses the Informer model for prediction because it uses multi-scale convolution operations, time series attention mechanisms, and adaptive feature selection mechanisms. These features can effectively extract and utilize multi-scale features from high-frequency signal components and automatically learn the spatiotemporal dependencies of time series data. In addition, the Transformer structure in the Informer model can capture the long-term trends and periodic change characteristics in high-frequency signal components, thereby better adapting to the characteristics of these components.

[0069] S6. Input the high-frequency signal component and the low-frequency signal component into a time series prediction model to obtain a prediction result of each high-frequency signal component and low-frequency signal component, and reconstruct the prediction result of each high-frequency signal component and low-frequency signal component to obtain a prediction result of the short-term power load mixture.

[0070] The embodiment of the present invention further includes: performing a comparative experiment on the hybrid short-term power load forecasting result and the comparison model to analyze the prediction performance of different models. Figure 3 Shown is an overall flow chart of the time series prediction model of an embodiment of the present invention.

[0071] A feature extraction method based on FSA-VMD-MIC is proposed. The FireflySparrow Algorithm (FSA) is used to find the optimal combination of influencing parameters in the variational mode decomposition (VMD) to obtain the best signal component. Since the signal components contain different influencing characteristics and time series information, the maximum information coefficient (MIC) is used to screen the features of each signal component and establish a feature matrix.

[0072] A hybrid prediction model based on LSTM-Informer is proposed. The zero-crossing rate is used as an indicator to determine the demarcation point of high-frequency and low-frequency signals. The Informer model is used to predict the high-frequency signal component, and the LSTM is used to predict the low-frequency signal component. All prediction results are reconstructed to obtain the final prediction value, which can effectively improve the prediction accuracy and is conducive to the safe and stable operation of the power system.

[0073] By adopting the embodiments of the present invention, the following beneficial effects are achieved:

[0074] The embodiment of the present invention combines the characteristics of the five methods of FSA, VMD, MIC, LSTM and Informer, gives full play to their respective advantages, can significantly improve the prediction accuracy, and meet the needs of refined prediction of power system load.

[0075] System Example

[0076] According to an embodiment of the present invention, a hybrid short-term power load forecasting system is provided. Figure 4 is a schematic diagram of a hybrid power short-term load forecasting system according to an embodiment of the present invention. Figure 4 As shown, the hybrid short-term power load forecasting system of the embodiment of the present invention specifically includes:

[0077] A data collection module 41, for collecting relevant power load and meteorological data as a first data set;

[0078] An optimal parameter acquisition module 42 is used to acquire an optimal combination of the mode number K and the penalty parameter α of the signal decomposition algorithm by using a preset optimization algorithm;

[0079] A signal component acquisition module 43, which uses a signal decomposition algorithm based on an optimal combination to decompose the first data set into a plurality of signal components, wherein the signal components have different frequencies and different amplitudes;

[0080] The input matrix acquisition module 44 calculates the maximum mutual information coefficient of each signal component using the relevant performance index, and selects the most relevant features of a preset number of signal components according to the maximum mutual information coefficient of each signal component to construct an input matrix;

[0081] A high- and low-frequency signal separation module 45 is used to determine a zero-crossing rate separation point using an adaptive algorithm according to the characteristic information in the input matrix, and to separate the high and low frequencies of the signal components according to the zero-crossing rate separation point to obtain a high-frequency signal component and a low-frequency signal component;

[0082] The result prediction module 46 is used to input the high-frequency signal component and the low-frequency signal component into the time series prediction model to obtain the prediction results of each high-frequency signal component and the low-frequency signal component, and reconstruct the prediction results of each high-frequency signal component and the low-frequency signal component to obtain the prediction results of the short-term power load mixture.

[0083] Device Example 1

[0084] According to an embodiment of the present invention, there is provided an electronic device, including:

[0085] processor; and,

[0086] A memory arranged to store computer executable instructions, which, when executed, cause the processor to perform the steps described in the above method embodiment.

[0087] Device Example 2

[0088] According to an embodiment of the present invention, a storage medium is provided for storing computer executable instructions, wherein the computer executable instructions, when executed, implement the steps described in the above method embodiment.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A hybrid short-term power load forecasting method, characterized in that include: S1. Collect relevant power load and meteorological data as a first data set; S2. Using a preset optimization algorithm to obtain an optimal combination of the modal number K and the penalty parameter α of the signal decomposition algorithm; S3, decomposing the first data set into a plurality of signal components using a signal decomposition algorithm based on an optimal combination, wherein the signal components have different frequencies and different amplitudes; S4, using the correlation performance index to calculate the maximum mutual information coefficient of each signal component, and selecting the most relevant features of a preset number of signal components according to the maximum mutual information coefficient of each signal component to construct an input matrix; S5, using an adaptive algorithm to determine a zero-crossing rate separation point according to the characteristic information in the input matrix, and separating the high and low frequencies of the signal components according to the zero-crossing rate separation point to obtain a high-frequency signal component and a low-frequency signal component; S6. Input the high-frequency signal component and the low-frequency signal component into a time series prediction model to obtain a prediction result of each high-frequency signal component and low-frequency signal component, and reconstruct the prediction result of each high-frequency signal component and low-frequency signal component to obtain a prediction result of the short-term power load mixture.

2. The method according to claim 1, characterized in that The hybrid short-term power load forecasting method further comprises: A comparative experiment is conducted between the prediction results of the mixed short-term power load and the prediction results of the comparison model to analyze the prediction performance of the time series prediction model and the comparison model.

3. The method according to claim 1, characterized in that The S2 specifically includes: The firefly algorithm and sparrow search algorithm are combined to obtain the optimal combination.

4. The method according to claim 3, characterized in that: The method of combining the firefly algorithm and the sparrow search algorithm to obtain the optimal combination specifically includes: S21, randomly generate one or more parameter combinations [K, α] of modal number K and penalty parameter α as the initial positions of the discoverer and the follower; S22, performing VMD decomposition on the initial positions of the finder and the follower, and calculating the fitness values ​​of the finder and the follower according to the decomposition results; S23, updating the positions of the discoverer and the follower based on the alarm so that the discoverer and the follower escape from the dangerous area, and updating the positions of the sparrows based on the fitness values ​​of the discoverer and the follower obtained in S22; S24, interfering with the sparrow position by the firefly algorithm to obtain a second sparrow position, if the second sparrow position is better than the sparrow position obtained in S23, then using the second sparrow position as a better sparrow position, otherwise, using the sparrow position obtained in S23 as the better sparrow position; S25, using a better sparrow position to perform VMD decomposition on the updated finder and follower positions, and calculating the fitness values ​​of the finder and follower according to each decomposition result; S26. Randomly select alarms and iteratively update the positions of the discoverers and followers. Update the sparrow positions based on the fitness values ​​of the discoverers and followers obtained in S25. Determine whether the fitness values ​​of the discoverers and followers obtained in S25 meet the preset stop conditions. If so, obtain the minimum fitness value among the current fitness values ​​of the discoverers and followers, and obtain the modal number K and penalty parameter α corresponding to the minimum fitness value as the optimal combination [K, α]. Otherwise, execute step S23.

5. A hybrid short-term power load forecasting system, characterized in that include: A data collection module, used for collecting relevant power load and meteorological data as a first data set; An optimal parameter acquisition module is used to obtain an optimal combination of the mode number K and the penalty parameter α of the signal decomposition algorithm using a preset optimization algorithm; A signal component acquisition module is configured to decompose the first data set into a plurality of signal components by using a signal decomposition algorithm based on an optimal combination, wherein the signal components have different frequencies and different amplitudes; Obtaining an input matrix module, calculating the maximum mutual information coefficient of each signal component using the relevant performance index, and selecting the most relevant features of a preset number of signal components according to the maximum mutual information coefficient of each signal component to construct an input matrix; A high- and low-frequency signal separation module is used to determine a zero-crossing rate separation point using an adaptive algorithm according to the characteristic information in the input matrix, and to separate the high and low frequencies of the signal components according to the zero-crossing rate separation point to obtain a high-frequency signal component and a low-frequency signal component; The result prediction module is used to input the high-frequency signal component and the low-frequency signal component into the time series prediction model to obtain the prediction result of each high-frequency signal component and the low-frequency signal component, and reconstruct the prediction result of each high-frequency signal component and the low-frequency signal component to obtain the prediction result of the short-term power load mixture.

6. The system according to claim 5, characterized in that The electric power short-term load hybrid forecasting system further comprises: The model comparison module is used to conduct a comparative experiment on the prediction result of the mixed short-term power load and the prediction result of the comparison model, and analyze the prediction performance of the time series prediction model and the comparison model.

7. The system according to claim 5, characterized in that The module for obtaining the optimal parameters adopts a combination of the firefly algorithm and the sparrow search algorithm to obtain the optimal combination.

8. The system according to claim 5, characterized in that The specific steps of obtaining the optimal combination by the optimal parameter obtaining module include: S21, randomly generate one or more parameter combinations [K, α] of modal number K and penalty parameter α as the initial positions of the discoverer and the follower; S22, performing VMD decomposition on the initial positions of the finder and the follower, and calculating the fitness values ​​of the finder and the follower according to the decomposition results; S23, updating the positions of the discoverer and the follower based on the alarm so that the discoverer and the follower escape from the dangerous area, and updating the positions of the sparrows based on the fitness values ​​of the discoverer and the follower obtained in S22; S24, interfering with the sparrow position by the firefly algorithm to obtain a second sparrow position, if the second sparrow position is better than the sparrow position obtained in S23, then using the second sparrow position as a better sparrow position, otherwise, using the sparrow position obtained in S23 as the better sparrow position; S25, using a better sparrow position to perform VMD decomposition on the updated finder and follower positions, and calculating the fitness values ​​of the finder and follower according to each decomposition result; S26. Randomly select alarms and iteratively update the positions of the discoverers and followers. Update the sparrow positions based on the fitness values ​​of the discoverers and followers obtained in S25. Determine whether the fitness values ​​of the discoverers and followers obtained in S25 meet the preset stop conditions. If so, obtain the minimum fitness value among the current fitness values ​​of the discoverers and followers, and obtain the modal number K and penalty parameter α corresponding to the minimum fitness value as the optimal combination [K, α]. Otherwise, execute step S23.

9. An electronic device, comprising: processor; as well as, A memory arranged to store computer executable instructions, which, when executed, cause the processor to perform the steps of the power short-term load hybrid forecasting method as claimed in any one of claims 1 to 4.

10. A storage medium for storing computer executable instructions, wherein the computer executable instructions, when executed, implement the steps of the method for hybrid short-term power load forecasting as claimed in any one of claims 1 to 4.

Images