A method and system for predicting power load

Abstract

The present invention discloses a method and system for predicting electric load, the method comprising: collecting historical load data of a target power grid system, constructing a historical load data matrix, obtaining a corresponding low-dimensional load data matrix, obtaining each initial sparse matrix that meets the preset normal distribution requirements, constructing an objective function, and performing iterative gradient sparsification processing on each initial sparse matrix based on the objective function to obtain an updated sparse matrix; performing element elimination and reconstruction processing on the historical load data matrix, the low-dimensional load data matrix, and the updated sparse matrix to obtain each error matrix; performing singular recombination processing on the updated sparse matrix, the low-dimensional load data matrix, and all error matrices to obtain a load recombination update matrix and a sparse recombination update matrix, and then constructing a load prediction sample data set to predict the electric load of the target power grid system. A method and system for predicting electric load provided by an embodiment of the present invention improves the prediction accuracy of electric load data.

Description

Technical Field

[0001] The present invention relates to the technical field of power load forecasting, and in particular to a power load forecasting method and system. Background Art

[0002] Power load forecasting refers to the process of predicting and estimating the power demand or power consumption in the future. Power load forecasting plays a vital role in the planning, operation and control of power systems. Accurate power load forecasting results help power companies to reasonably arrange production plans, optimize resource allocation, and improve power supply reliability. At the same time, power load forecasting is also one of the important foundations for the construction of smart grids, which helps to promote the intelligent development of power systems.

[0003] However, the existing power load forecasting method predicts the power load based on directly acquired sample data, and the power load forecasting result obtained by this method has low accuracy.

[0004] It can be seen that how to improve the prediction accuracy of power load has become a technical problem that technical personnel in this field need to solve urgently. Summary of the invention

[0005] The present invention provides a power load forecasting method and system to solve the technical problem that the existing power load forecasting method predicts the power load based on directly acquired sample data, resulting in low accuracy of the obtained power load forecasting results.

[0006] In order to solve the above technical problems, an embodiment of the present invention provides a method for predicting power load.

[0007] Collect historical load data of the target power grid system, and obtain a historical load data matrix based on the historical load data; perform singular screening on the historical load data matrix to obtain a corresponding low-dimensional load data matrix; and obtain each initial sparse matrix that meets the preset normal distribution requirements;

[0008] Performing linear regularization processing on the historical load data matrix, the low-dimensional load data matrix and each of the initial sparse matrices to construct an objective function; performing iterative gradient sparse processing on each of the initial sparse matrices based on the objective function to obtain an updated sparse matrix;

[0009] Performing element elimination and reconstruction processing on the historical load data matrix, the low-dimensional load data matrix and the updated sparse matrix to obtain respective error matrices; performing singular reorganization processing on the updated sparse matrix, the low-dimensional load data matrix and all the error matrices to obtain a load reorganization update matrix and a sparse reorganization update matrix;

[0010] A load prediction sample data set is constructed based on the linear relationship between the sparse reorganization update matrix and the load reorganization update matrix; and the power load of the target power grid system is predicted using the load prediction sample data set.

[0011] As one of the preferred solutions, the iterative gradient sparse processing is performed on each of the initial sparse matrices based on the objective function to obtain an updated sparse matrix, including:

[0012] Iteratively perform gradient soft threshold processing on the objective function, use the result of each gradient soft threshold processing as the input of the initial sparse matrix of the next gradient soft threshold processing, so as to update the initial sparse matrix until the initial sparse matrix converges, and use the final result of the gradient soft threshold processing as the updated sparse matrix.

[0013] As one of the preferred solutions, the expression of the objective function is:

[0014]

[0015] in, is the objective function, represents the minimum value function, is the initial sparse matrix, is the historical load data matrix, is the low-dimensional load data matrix, The historical load data matrix and the corresponding sparse representation The reconstruction error between To preset the regularization parameter, A tree-structured sparsity constraint representing the initial sparse matrix.

[0016] As one of the preferred solutions, the expression of the gradient soft threshold processing is:

[0017]

[0018]

[0019] in, is the initial sparse matrix of the k+1th gradient soft threshold processing, represents the soft threshold operation, is the initial sparse matrix of the k-th gradient soft threshold processing, is the preset step size, is the objective function of the k-th gradient soft threshold processing;

[0020] represents constrained gradient computation, is the low-dimensional load data matrix, is the transpose of the low-dimensional load data matrix, is the historical load data matrix, is the symbolic function, is the preset regularization parameter.

[0021] As one of the preferred solutions, the historical load data matrix, the low-dimensional load data matrix and the updated sparse matrix are subjected to element elimination and reconstruction processing to obtain respective error matrices, including:

[0022]

[0023] in, is the kth error matrix, X is the historical load data matrix, , is the number of rows of the low-dimensional load data matrix, is the reconstructed load matrix obtained by removing the j-th row of the low-dimensional load data matrix, The reconstructed sparse matrix is ​​obtained by removing the j-th column of the updated sparse matrix.

[0024] As one of the preferred solutions, the updating sparse matrix, the low-dimensional load data matrix and all the error matrices are subjected to singular reorganization processing to obtain a load reorganization update matrix and a sparse reorganization update matrix, including:

[0025] Screening and analyzing all row vectors in the updated sparse matrix to obtain various optional vector indexes;

[0026] Concatenate the column vectors corresponding to all the optional vector indices in each of the error matrices to obtain a corresponding reconstruction error matrix;

[0027] The update sparse matrix and the low-dimensional load data matrix are sparsely reconstructed according to the singular value decomposition results of all the reconstruction error matrices to obtain a load reorganization update matrix and a sparse reorganization update matrix.

[0028] As one of the preferred solutions, the update sparse matrix and the low-dimensional load data matrix are sparsely reconstructed according to the singular value decomposition results of all the reconstruction error matrices to obtain a load reorganization update matrix and a sparse reorganization update matrix, including:

[0029] Performing singular value decomposition on each of the reconstruction error matrices to obtain a corresponding reconstruction error left singular matrix, a corresponding reconstruction error singular value matrix, and a corresponding reconstruction error right singular matrix;

[0030] Taking the product of the first row vector of each reconstruction error right singular matrix and the first singular value of the corresponding reconstruction error singular value matrix as the singular eigenvector of each reconstruction error right singular matrix;

[0031] Each column vector of the low-dimensional load data matrix is ​​replaced by the first column vector of the corresponding left singular matrix of the reconstruction error to obtain a load reorganization update matrix; each row vector of the update sparse matrix is ​​replaced by the singular eigenvector of the corresponding right singular matrix of the reconstruction error to obtain a sparse reorganization update matrix.

[0032] As one of the preferred solutions, the singular screening process is performed on the historical load data matrix to obtain the corresponding low-dimensional load data matrix, including:

[0033] Performing singular value decomposition on the historical load data matrix to obtain a corresponding historical load left singular matrix, a corresponding historical load singular value matrix, and a corresponding historical load right singular matrix;

[0034] All column vectors of the left singular matrix of the historical load are sorted and screened, and a low-dimensional load data matrix is ​​constructed based on the results of the sorting and screening.

[0035] As one of the preferred solutions, the method for obtaining the initial sparse matrix is:

[0036] Randomly generate a number of data points that meet the preset normal distribution requirements, and matrix-process all the data points to obtain an initial sparse matrix;

[0037] The number of rows of the initial sparse matrix is ​​equal to the number of columns of the low-dimensional load data matrix.

[0038] Another embodiment of the present invention provides an electric load forecasting system, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements an electric load forecasting method as described above.

[0039] Compared with the prior art, the embodiments of the present invention have the following advantages:

[0040] (1) Considering that the existing power load forecasting method is based on directly obtained sample data to predict the power load, the power load forecasting results obtained by this method are of low accuracy. In order to improve the accuracy of the power load forecasting results, a historical load data matrix is ​​obtained based on the historical load data of the target power grid system collected, and the historical load data matrix is ​​subjected to singular screening to obtain the corresponding low-dimensional load data matrix. By screening the historical load data and constructing a low-dimensional load data matrix, the redundant information in the historical load data is preliminarily eliminated, thereby improving the subsequent prediction accuracy of the power load;

[0041] (2) Obtain each initial sparse matrix that meets the preset normal distribution requirements, perform linear regularization processing on the historical load data matrix, the low-dimensional load data matrix and each initial sparse matrix, construct an objective function, and perform iterative gradient sparsification processing on each initial sparse matrix based on the objective function to obtain an updated sparse matrix. Through the gradient calculation and soft threshold operation of the reconstruction error and regularization constraint of the historical load data matrix, the updated sparse matrix is ​​made sparse in the tree structure, further eliminating redundant information in the historical load data, thereby improving the subsequent prediction accuracy of the power load;

[0042] (3) Considering that load data will be continuously updated as time and conditions change, in order to reflect these changes and ensure the accuracy of power load data prediction, the historical load data matrix, low-dimensional load data matrix and updated sparse matrix are reconstructed by element elimination to obtain various error matrices. The updated sparse matrix, low-dimensional load data matrix and all error matrices are singularly reorganized to obtain load reorganization update matrix and sparse reorganization update matrix, so as to update the sparse matrix and load data in real time, ensuring that after the load data is updated, the values ​​in the sparse matrix can also correctly reflect the sparse representation of the load data. Based on the linear relationship between the sparse reorganization update matrix and the load reorganization update matrix, a load prediction sample data set is constructed, which improves the real-time prediction of power load.

[0043] (4) By constructing a low-dimensional load data matrix and an initial sparse matrix based on historical load data, the soft threshold operation and singular value decomposition and reconstruction are used for sparse processing, redundant information is eliminated, and real-time updates are performed to obtain a load forecasting sample data set, thereby comprehensively improving the prediction accuracy of power load data. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A flowchart of a method for predicting power load in one embodiment of the present invention;

[0045] Figure 2 It is a schematic diagram of obtaining an error matrix in one embodiment of the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] In the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and are not to be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first", "second", "third", etc. may explicitly or implicitly include one or more of the feature. In the description of this application, unless otherwise specified, "plurality" means two or more.

[0048] In the description of the present application, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be a connection between the two elements. The terms "vertical", "horizontal", "left", "right", "upper", "lower" and similar expressions used herein are only for illustrative purposes, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood according to specific circumstances.

[0049] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as those commonly understood by those skilled in the art. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in this application can be understood by specific circumstances.

[0050] An embodiment of the present invention provides a method for predicting power load. For details, see Figure 1~Figure 2 , Figure 1 FIG. 1 is a flow chart of a method for predicting power load in one embodiment of the present invention. Figure 2FIG. 4 is a schematic diagram showing how to obtain an error matrix in one embodiment of the present invention.

[0051] A flowchart of a power load forecasting method in one embodiment of the present invention includes the following steps S1 to S4, which are as follows:

[0052] Step S1: Collect historical load data of the target power grid system, and obtain a historical load data matrix based on the historical load data; perform singular screening on the historical load data matrix to obtain a corresponding low-dimensional load data matrix; and obtain each initial sparse matrix that meets the preset normal distribution requirements.

[0053] It should be noted that the existing power load forecasting method predicts the power load based on the directly acquired sample data. The power load forecasting results obtained by this method have low accuracy. In order to improve the accuracy of the power load forecasting results, the sample data needs to be processed first.

[0054] Specifically, historical load data of the target power grid system is collected, and all historical load data are organized and processed in the form of a matrix to obtain a historical load data matrix.

[0055] Furthermore, the historical load data matrix is ​​subjected to singular value decomposition to obtain the corresponding historical load left singular matrix, the corresponding historical load singular value matrix and the corresponding historical load right singular matrix; all column vectors of the historical load left singular matrix are sorted and screened, and a low-dimensional load data matrix is ​​constructed based on the results of the sorting and screening.

[0056] It should be noted that singular value decomposition can effectively reduce the dimension of a data set while retaining the most important information by retaining the features corresponding to the largest singular value. Singular value decomposition is a well-known technology and will not be described in detail in this embodiment.

[0057] As an embodiment of the present application, a matrix formed by arranging the first K column vectors of the historical load left singular matrix according to the number of columns is used as a low-dimensional load data matrix.

[0058] A number of data points that meet the preset normal distribution requirements are randomly generated, and all the data points are matrixed to obtain an initial sparse matrix; wherein the number of rows of the initial sparse matrix is ​​equal to the number of columns of the low-dimensional load data matrix.

[0059] It should be noted that a certain degree of randomness is introduced through the normal distribution to avoid falling into the local optimum too early. The role of matrix processing is to organize all data points in the form of a matrix.

[0060] Step S2: Perform linear regularization processing on the historical load data matrix, the low-dimensional load data matrix and each initial sparse matrix to construct an objective function; perform iterative gradient sparse processing on each initial sparse matrix based on the objective function to obtain an updated sparse matrix.

[0061] Specifically, the expression of the objective function is:

[0062]

[0063] in, is the objective function, represents the minimum value function, is the initial sparse matrix, is the historical load data matrix, is the low-dimensional load data matrix, is the historical load data matrix and the corresponding sparse representation The reconstruction error between To preset the regularization parameter, Represents a tree-based sparsity constraint on an initially sparse matrix.

[0064] It should be noted that the tree-structure-based sparsity constraint causes the initial sparse matrix to be sparse in the tree structure, and the preset regularization parameter is an artificially preset value used to control the intensity of sparsity.

[0065] Furthermore, the objective function is iteratively subjected to gradient soft threshold processing, and the result of each gradient soft threshold processing is used as the initial sparse matrix for the next gradient soft threshold processing to update the initial sparse matrix until the initial sparse matrix converges, and the final result of the gradient soft threshold processing is used as the updated sparse matrix.

[0066] In this step, the expression of gradient soft threshold processing is:

[0067]

[0068]

[0069] in, is the initial sparse matrix for the k+1th gradient soft threshold processing, represents the soft threshold operation, is the initial sparse matrix for the kth gradient soft threshold processing, is the preset step size, is the objective function of the k-th gradient soft threshold processing;

[0070] represents constrained gradient computation, is the low-dimensional load data matrix, is the transpose of the low-dimensional loading data matrix, is the historical load data matrix, is the symbolic function, is the preset regularization parameter.

[0071] It should be noted that the soft threshold operation is a nonlinear signal processing technology. It is based on the assumption of signal sparsity. By comparing the amplitude of the signal and applying a specific processing method, the soft threshold operation can effectively suppress the noise in the signal while retaining the main features of the signal. The gradient soft threshold processing allows the initial sparse matrix to be gradually adjusted to a smaller reconstruction error and a stronger sparsity direction through gradient calculation. The sign function is a special function that is used to return an integer variable that represents the positive and negative signs of the input parameters. The calculation of the soft threshold operation and the sign function are both well-known technologies, and this embodiment will not be repeated here.

[0072] Step S3: Perform element elimination and reconstruction processing on the historical load data matrix, the low-dimensional load data matrix and the updated sparse matrix to obtain various error matrices; perform singular recombination processing on the updated sparse matrix, the low-dimensional load data matrix and all error matrices to obtain the load recombination update matrix and the sparse recombination update matrix.

[0073] It should be noted that as time and conditions change, load data will be continuously updated. In order to reflect these changes and ensure the accuracy of the data, the load data needs to be iteratively updated.

[0074] Specifically, the historical load data matrix, the low-dimensional load data matrix and the updated sparse matrix are subjected to element elimination and reconstruction processing to obtain various error matrices. The expression of element elimination and reconstruction processing is:

[0075]

[0076] in, is the kth error matrix, X is the historical load data matrix, , is the number of rows in the low-dimensional loading data matrix, is the reconstructed dictionary obtained by removing the jth row of the low-dimensional load data matrix, The reconstructed sparse matrix is ​​obtained by eliminating the j-th column of the updated sparse matrix.

[0077] Furthermore, all row vectors in the updated sparse matrix are screened and analyzed to obtain the various optional vector indices, and the column vectors corresponding to all the optional vector indices in each error matrix are concatenated to obtain the corresponding reconstructed error matrix. The updated sparse matrix and the low-dimensional load data matrix are sparsely reconstructed according to the singular value decomposition results of all the reconstructed error matrices to obtain the reorganization update dictionary and the sparse reorganization update matrix, thereby correcting the initial sparse matrix to ensure that after the load data is updated, the values ​​in the sparse matrix can also correctly reflect the sparse representation of the load data.

[0078] In this step, the update sparse matrix and the low-dimensional load data matrix are sparsely reconstructed according to the singular value decomposition results of all reconstruction error matrices to obtain a reorganization update dictionary and a sparse reorganization update matrix, including:

[0079] Perform singular value decomposition on each reconstruction error matrix to obtain the corresponding reconstruction error left singular matrix, the corresponding reconstruction error singular value matrix and the corresponding reconstruction error right singular matrix; take the product of the first row vector of each reconstruction error right singular matrix and the first singular value of the corresponding reconstruction error singular value matrix as the singular eigenvector of each reconstruction error right singular matrix; replace each column vector of the low-dimensional load data matrix with the first column vector of the corresponding reconstruction error left singular matrix to obtain a reorganization update dictionary; replace each row vector of the update sparse matrix with the singular eigenvector of the corresponding reconstruction error right singular matrix to obtain a sparse reorganization update matrix.

[0080] Step S4: construct a load prediction sample data set based on the linear relationship between the sparse reorganization update matrix and the load reorganization update matrix; and predict the power load of the target power grid system using the load prediction sample data set.

[0081] As an embodiment of the present application, a sparse reorganization update matrix and a load reorganization update matrix are multiplied to obtain a load prediction sample data set.

[0082] Furthermore, based on the load forecasting sample data set, the long short-term memory network model is used to predict the power load and obtain the power load forecasting data of the target power grid system.

[0083] It should be noted that the long short-term memory network model is a well-known technology and will not be elaborated in this application.

[0084] Another embodiment of the present invention provides an electric load forecasting system, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, it implements an electric load forecasting method as described above.

[0085] The power load forecasting method and system provided by the embodiment of the present invention have the following beneficial effects compared with the prior art:

[0086] (1) Considering that the existing power load forecasting method is based on directly obtained sample data to predict the power load, the power load forecasting results obtained by this method are of low accuracy. In order to improve the accuracy of the power load forecasting results, a historical load data matrix is ​​obtained based on the historical load data of the target power grid system collected, and the historical load data matrix is ​​subjected to singular screening to obtain the corresponding low-dimensional load data matrix. By screening the historical load data and constructing a low-dimensional load data matrix, the redundant information in the historical load data is preliminarily eliminated, thereby improving the subsequent prediction accuracy of the power load;

[0087] (2) Obtain each initial sparse matrix that meets the preset normal distribution requirements, perform linear regularization processing on the historical load data matrix, the low-dimensional load data matrix and each initial sparse matrix, construct an objective function, and perform iterative gradient sparsification processing on each initial sparse matrix based on the objective function to obtain an updated sparse matrix. Through the gradient calculation and soft threshold operation of the reconstruction error and regularization constraint of the historical load data matrix, the updated sparse matrix is ​​made sparse in the tree structure, further eliminating redundant information in the historical load data, thereby improving the subsequent prediction accuracy of the power load;

[0088] (3) Considering that load data will be continuously updated as time and conditions change, in order to reflect these changes and ensure the accuracy of power load data prediction, the historical load data matrix, low-dimensional load data matrix and updated sparse matrix are reconstructed by element elimination to obtain various error matrices. The updated sparse matrix, low-dimensional load data matrix and all error matrices are singularly reorganized to obtain load reorganization update matrix and sparse reorganization update matrix, so as to update the sparse matrix and load data in real time, ensuring that after the load data is updated, the values ​​in the sparse matrix can also correctly reflect the sparse representation of the load data. Based on the linear relationship between the sparse reorganization update matrix and the load reorganization update matrix, a load prediction sample data set is constructed, which improves the real-time prediction of power load.

[0089] (4) By constructing a low-dimensional load data matrix and an initial sparse matrix based on historical load data, the soft threshold operation and singular value decomposition and reconstruction are used for sparse processing, redundant information is eliminated, and real-time updates are performed to obtain a load forecasting sample data set, thereby comprehensively improving the prediction accuracy of power load data.

[0090] The above-mentioned embodiments only express several implementation methods of the present invention, and the description thereof is relatively specific and detailed, but it cannot be understood as limiting the scope of the patent of the present invention. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, which all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A method for predicting power load, characterized in that: The method comprises: Collect historical load data of the target power grid system, and obtain a historical load data matrix based on the historical load data; perform singular screening on the historical load data matrix to obtain a corresponding low-dimensional load data matrix; and obtain each initial sparse matrix that meets the preset normal distribution requirements; The historical load data matrix, the low-dimensional load data matrix and each of the initial sparse matrices are subjected to linear regularization processing to construct an objective function; wherein the expression of the objective function is: in, is the objective function, represents the minimum value function, is the initial sparse matrix, is the historical load data matrix, is the low-dimensional load data matrix, The historical load data matrix and the corresponding sparse representation The reconstruction error between To preset the regularization parameter, Representing a tree-structured sparsity constraint of the initial sparse matrix; Based on the objective function, each of the initial sparse matrices is subjected to iterative gradient sparse processing to obtain an updated sparse matrix; wherein, based on the objective function, each of the initial sparse matrices is subjected to iterative gradient sparse processing to obtain an updated sparse matrix, including: Iteratively perform gradient soft threshold processing on the objective function, use the result of each gradient soft threshold processing as the input of the initial sparse matrix of the next gradient soft threshold processing, so as to update the initial sparse matrix until the initial sparse matrix converges, and use the final result of the gradient soft threshold processing as the updated sparse matrix; wherein the expression of the gradient soft threshold processing is: in, is the initial sparse matrix of the k+1th gradient soft threshold processing, represents the soft threshold operation, is the initial sparse matrix of the k-th gradient soft threshold processing, is the preset step size, is the objective function of the k-th gradient soft threshold processing; represents constrained gradient computation, is the low-dimensional load data matrix, is the transpose of the low-dimensional load data matrix, is the historical load data matrix, is the symbolic function, is the preset regularization parameter; The historical load data matrix, the low-dimensional load data matrix and the updated sparse matrix are subjected to element elimination and reconstruction processing to obtain respective error matrices; wherein, the historical load data matrix, the low-dimensional load data matrix and the updated sparse matrix are subjected to element elimination and reconstruction processing to obtain respective error matrices, including: in, is the kth error matrix, X is the historical load data matrix, , is the number of rows of the low-dimensional load data matrix, is the reconstructed load matrix obtained by removing the j-th row of the low-dimensional load data matrix, A reconstructed sparse matrix obtained by removing the j-th column of the updated sparse matrix; Performing singular reorganization processing on the update sparse matrix, the low-dimensional load data matrix and all the error matrices to obtain a load reorganization update matrix and a sparse reorganization update matrix; A load prediction sample data set is constructed based on the linear relationship between the sparse reorganization update matrix and the load reorganization update matrix; and the power load of the target power grid system is predicted using the load prediction sample data set.

2. A method for predicting power load according to claim 1, characterized in that: The step of performing singular reorganization processing on the update sparse matrix, the low-dimensional load data matrix and all the error matrices to obtain a load reorganization update matrix and a sparse reorganization update matrix comprises: Screening and analyzing all row vectors in the updated sparse matrix to obtain various optional vector indexes; Concatenate the column vectors corresponding to all the optional vector indices in each of the error matrices to obtain a corresponding reconstruction error matrix; The update sparse matrix and the low-dimensional load data matrix are sparsely reconstructed according to the singular value decomposition results of all the reconstruction error matrices to obtain a load reorganization update matrix and a sparse reorganization update matrix.

3. A method for predicting power load according to claim 2, characterized in that: The step of performing sparse reconstruction on the update sparse matrix and the low-dimensional load data matrix according to the singular value decomposition results of all the reconstruction error matrices to obtain a load reorganization update matrix and a sparse reorganization update matrix comprises: Performing singular value decomposition on each of the reconstruction error matrices to obtain a corresponding reconstruction error left singular matrix, a corresponding reconstruction error singular value matrix, and a corresponding reconstruction error right singular matrix; Taking the product of the first row vector of each reconstruction error right singular matrix and the first singular value of the corresponding reconstruction error singular value matrix as the singular eigenvector of each reconstruction error right singular matrix; Each column vector of the low-dimensional load data matrix is ​​replaced by the first column vector of the corresponding left singular matrix of the reconstruction error to obtain a load reorganization update matrix; each row vector of the update sparse matrix is ​​replaced by the singular eigenvector of the corresponding right singular matrix of the reconstruction error to obtain a sparse reorganization update matrix.

4. A method for predicting power load according to claim 1, characterized in that: The performing singular screening processing on the historical load data matrix to obtain a corresponding low-dimensional load data matrix includes: Performing singular value decomposition on the historical load data matrix to obtain a corresponding historical load left singular matrix, a corresponding historical load singular value matrix, and a corresponding historical load right singular matrix; All column vectors of the left singular matrix of the historical load are sorted and screened, and a low-dimensional load data matrix is ​​constructed based on the results of the sorting and screening.

5. A method for predicting power load according to claim 1, characterized in that: The method for obtaining the initial sparse matrix is: Randomly generate a number of data points that meet the preset normal distribution requirements, and matrix-process all the data points to obtain an initial sparse matrix; The number of rows of the initial sparse matrix is ​​equal to the number of columns of the low-dimensional load data matrix.

6. A power load forecasting system, characterized in that: The system includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements a power load forecasting method as described in any one of claims 1 to 5 when executing the computer program.

Images