Transformer area multi-user short-term load prediction method and system based on space-time correlation

Abstract

The invention discloses a transformer area multi-user short-term load prediction method and system based on spatial-temporal correlation, and relates to the technical field of power load prediction. Comprising the following steps: identifying abnormal load data by adopting a DBSCAN algorithm, clustering users through a K-means algorithm, and replacing abnormal characteristic values with mean values in classes at the same time to obtain an optimized load data set; constructing a comprehensive correlation analysis index to carry out quantitative analysis on the correlation between each influence factor and the load, and selecting the influence factors with strong correlation as input characteristics of the multi-element load prediction model; characteristic extraction and prediction are respectively carried out in time and space by using Informer and GCN, and then single prediction results are fused through a variance-covariance combined prediction model to realize load prediction in a certain period of time in the future. According to the method, the short-term load prediction precision can be effectively improved by means of the introduced space-time related information of the transformer area multi-user load.

Description

Technical Field

[0001] The present invention relates to the technical field of power load forecasting, and more specifically to a method and system for forecasting short-term load of multiple users in an area based on time-space correlation. Background Art

[0002] Accurate short-term load forecasting is crucial for the safe and stable operation of power systems as well as for optimal control and scheduling. Scholars at home and abroad have been committed to studying better short-term load forecasting methods. Existing research mainly includes traditional time series analysis methods, machine learning-based methods, and deep learning-based methods.

[0003] However, with the massive access of clean energy to the power grid and the increasing number of electric vehicles, the complexity and uncertainty of the power grid are becoming increasingly serious. In addition, the widespread use of devices such as smart meters on the user side has also brought about the demand for load forecasting for a large number of users, posing a challenge to the traditional power load forecasting model. Most of the existing methods model the temporal characteristics of a single user and ignore the spatial correlation between multiple users. However, the traditional prediction method that only targets a single object can no longer meet the new situation of massive users. Users in the same area share the same geographical space, meteorological conditions, electricity price policies and other comprehensive factors. There is often a certain spatiotemporal correlation between user electricity consumption behaviors. If this spatiotemporal correlation can be fully exploited, it will have a very positive significance for the modeling and prediction of short-term user loads.

[0004] Therefore, how to mine the spatiotemporal correlation of users and accurately predict the load of massive users is an urgent problem that technicians in this field need to solve. Summary of the invention

[0005] In view of this, the present invention provides a method and system for short-term load prediction of multiple users in a substation based on spatiotemporal correlation, which achieves accurate short-term prediction of multi-user load in a substation by introducing spatiotemporal correlation information of multi-user load in the substation.

[0006] In order to achieve the above object, the present invention provides the following technical solutions:

[0007] The present invention discloses a method for short-term load forecasting of multiple users in a substation area based on time-space correlation, and the specific steps are as follows:

[0008] Obtain the historical electricity load data set of multiple users in the substation area;

[0009] The DBSCAN algorithm is used to identify abnormal load data in the historical load data set, the users are clustered by the K-means algorithm, and the characteristic value of the abnormal load data is replaced by the mean value within the class to which the abnormal load data belongs at the same moment, so as to obtain an optimized load data set;

[0010] Using comprehensive correlation analysis indicators, quantitatively analyzing the correlation between each load influencing factor and the load in the optimized load data set, selecting influencing factors with strong correlation to obtain a predicted load data set as an input of a multivariate load prediction model;

[0011] The multivariate load forecasting model uses the Informer model and the GCN model to perform feature extraction and forecasting, and then fuses the forecast results of the Informer model and the GCN model through a variance-covariance combined forecasting model to obtain a final load forecasting result.

[0012] Furthermore, the specific steps of obtaining the optimized load data set are:

[0013] Step 1: In the historical load data set, the measured load data of all users in the substation area in the past t time period are intercepted to construct a sample matrix, and the sample matrix is ​​expressed as:

[0014]

[0015] Among them, P i,j represents the load measured by user i at the jth moment, P i represents a sample group of the sample matrix P, m represents the total number of users in the area, and k represents the number of sampling points;

[0016] Step 2: Use the K-means algorithm to divide all sample groups into c user groups a (a=1,2,…,c);

[0017] Step 3: Replace the characteristic value of the abnormal load data with the load average value in each user group at the same time.

[0018] Furthermore, the step 2 specifically includes:

[0019] Step 2.1: Randomly select c initial cluster centers P from all sample groups a ;

[0020] Step 2.2, calculate the Euclidean distance d from each sample group to each of the initial cluster centers, the formula is:

[0021]

[0022] Step 2.3: Any sample group P i Divide into the cluster center P that is closest to a In the class represented, traverse all sample groups and complete a division;

[0023] Step 2.4, calculate the sample mean of each class and calculate the sum of squared errors of the clusters:

[0024]

[0025] in, is the sample mean of the ath user group; N a is the total number of samples in the ath class; E is the sum of clustering square errors;

[0026] Step 2.5, taking the sample mean of each class as the new cluster center, returning to step 2.2 to perform iterative clustering division until the sum of cluster square errors converges to obtain the final clustering result.

[0027] Furthermore, the comprehensive correlation analysis index C M The formula is:

[0028]

[0029] Among them, ρ X,Y is the Pearson correlation coefficient, M IC is the maximum information coefficient value;

[0030] When C M When it is greater than the preset threshold σ, it means that there is a strong correlation between the influencing factor variable X and the load Y. The influencing factor with a strong correlation with the load forecast is selected as the input characteristic variable of the final load combination forecasting model.

[0031] Furthermore, the Pearson correlation coefficient value ρ X,Y The calculation formula is expressed as:

[0032]

[0033] Among them, Y i is the actual observed data sequence of the load, X i are characteristic data sequences that may affect load forecasting, l is the length of the sequence, and ρ X,Y It represents the Pearson correlation coefficient value between variable X and loading Y;

[0034] The maximum information coefficient value M IC The calculation formula is expressed as:

[0035]

[0036] Among them, I(X i ,Y j ) is X i With Y i The mutual information between |X i | and |Y i |respectively the sequence X i With Y i The length of min(|Xi |,|Y i |) means taking the sequence X i With Y i Minimum length.

[0037] Furthermore, the Informer model selects the load data of the user H time steps in the predicted load data set as input, predicts the load at the future time p, and obtains the predicted output result

[0038] Furthermore, the GCN model uses the weighted user adjacency matrix A and the user feature matrix B as input to predict the load at the next time p and obtain the output result

[0039] Furthermore, the weighted user adjacency matrix A is expressed as:

[0040] A={R i1 ,R i2 ,…,R if};

[0041] Among them, R if Represents the spatial correlation between user i and user f, which is calculated using the Pearson correlation coefficient method. The calculation formula is:

[0042]

[0043] Among them, P i is the actual observed data sequence of user i’s load; P f is the actual observed data sequence of user f’s load, l is the sequence length;

[0044] The user feature matrix B is expressed as:

[0045]

[0046] Among them, P i,j Represents the measured load value of user i at a certain time j in history.

[0047] Furthermore, the variance-covariance combined prediction model calculates the combined weight coefficients of the Informer model and the GCN model respectively, and outputs the final load forecast value at the next p moments according to the calculated combined weights; the variance-covariance combined prediction model is expressed as:

[0048]

[0049] in Respectively represent the sequence of load forecast values ​​predicted by the Informer model and the GCN model at the next p moments, y t represents the sequence of the final load forecast values, w T 、w S They are The weight coefficient of T +w S =1;

[0050] Weight coefficient w T 、w S The method for determining is:

[0051] Based on the predicted load data set, the corresponding prediction results are calculated using the Informer model, the GCN model and the variance-covariance combined prediction model. and t ;

[0052] The measured load data at time t is selected from the predicted load data set to form a sequence Calculate the error series of the three prediction results respectively:

[0053]

[0054] in, e t They are y t The corresponding error sequence;

[0055] The relationship between the weight coefficient and the error sequence is expressed as:

[0056]

[0057] The error variance of the final load forecast sequence is expressed as:

[0058]

[0059] Among them, Var(e t ) is the error variance of the final load forecast sequence, for covariance of w T Var(e t ) to find the minimum value, we get:

[0060]

[0061] w S =1-w T .

[0062] The present invention also discloses a short-term load forecasting system for multiple users in a substation area based on time-space correlation, comprising:

[0063] Data acquisition module: obtains the historical electricity load data set of multiple users in the substation area;

[0064] Data processing module: using DBSCAN algorithm to identify abnormal load data in the historical load data set, clustering users through K-means algorithm, replacing the characteristic value of abnormal load data with the mean value of the class to which the abnormal load data belongs at the same moment, and obtaining an optimized load data set;

[0065] Feature optimization module: using comprehensive correlation analysis indicators, quantitatively analyzing the correlation between each load influencing factor and the load in the optimized load data set, selecting influencing factors with strong correlation to obtain a predicted load data set as the input of the multivariate load prediction model;

[0066] Load forecasting module: The multivariate load forecasting model uses the Informer model and the GCN model to perform feature extraction and forecasting, and then fuses the forecast results of the Informer model and the GCN model through the variance-covariance combined forecasting model to obtain the final load forecasting result.

[0067] Through the above technical solutions, it can be known that compared with the prior art, the present invention discloses a method and system for short-term load prediction of multiple users in a substation based on spatiotemporal correlation, which can effectively improve the accuracy of short-term load prediction by relying on the introduced spatiotemporal correlation information of the load of multiple users in the substation. The abnormal data correction method based on K-means spatial clustering of the present invention uses the measured load as a feature to cluster users, and replaces the abnormal feature value with the mean value in the same moment class, which can effectively fuse spatiotemporal features to improve the accuracy of abnormal data correction; the correlation between each influencing factor and the load is quantitatively analyzed using a comprehensive correlation analysis index, and features with strong correlation are selected as the input of the model, which can effectively improve the prediction accuracy and efficiency while streamlining the model and data; Informer is used to extract time features, GCN is used to extract spatial features, and then the spatiotemporal features are fused through the variance-covariance combined prediction method to complete the final load prediction, which can make full use of spatiotemporal information to improve the accuracy of load prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present invention 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 embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0069] Figure 1 It is a schematic diagram of the overall process of an embodiment of the present invention;

[0070] Figure 2 This is a schematic diagram of the structure of the Informer model of an embodiment of the present invention;

[0071] Figure 3 Schematic diagram of the structure of the GCN model of an embodiment of the present invention;

[0072] Figure 4 The figure is a schematic diagram of the load combination forecasting process based on Informer and GCN according to an embodiment of the present invention. DETAILED DESCRIPTION

[0073] 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. 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.

[0074] The embodiment of the present invention discloses a method for short-term load forecasting of multiple users in a substation based on time-space correlation, such as Figure 1 As shown, the specific steps are as follows:

[0075] Obtain the historical electricity load data set of multiple users in the substation area;

[0076] The DBSCAN algorithm is used to identify abnormal load data in the historical load data set, and the users are clustered by the K-means algorithm. The characteristic value of the abnormal load data is replaced by the mean value of the class to which the abnormal load data belongs at the same moment, and the optimized load data set is obtained.

[0077] The comprehensive correlation analysis index is used to quantitatively analyze the correlation between each load influencing factor and the load in the optimized load data set, and the influencing factors with strong correlation are selected to obtain the predicted load data set as the input of the multivariate load prediction model;

[0078] The multivariate load forecasting model uses the Informer model and the GCN model for feature extraction and prediction, and then integrates the prediction results of the Informer model and the GCN model through the variance-covariance combined prediction model to obtain the final load forecasting result.

[0079] In a specific embodiment, the specific steps of obtaining the optimized load data set are:

[0080] Step 1: In the historical load data set, the measured load data of all users in the substation area in the past t time period are intercepted to construct a sample matrix. The sample matrix is ​​expressed as:

[0081]

[0082] Among them, P i,j represents the load measured by user i at the jth moment, P i represents a sample group of the sample matrix P, m represents the total number of users in the station area, and k represents the number of sampling points.

[0083] Step 2: Use the K-means algorithm to divide all sample groups into c user groups a (a=1,2,…,c);

[0084] Step 3: Replace the characteristic value of the abnormal load data with the load mean value in each user group at the same time.

[0085] In a specific embodiment, step 2 specifically includes:

[0086] Step 2.1: Randomly select c initial cluster centers P from all sample groups a ;

[0087] Step 2.2, calculate the Euclidean distance d from each sample group to each initial cluster center, the formula is:

[0088]

[0089] Step 2.3: Any sample group P i Divide into the nearest cluster center P a In the class represented, traverse all sample groups and complete a division;

[0090] Step 2.4, calculate the sample mean of each class and calculate the sum of squared errors of the clusters:

[0091]

[0092] in, is the sample mean of the ath user group; N a is the total number of samples in the ath class; E is the sum of clustering square errors;

[0093] Step 2.5, take the sample mean of each class as the new cluster center, return to step 2.2 for iterative clustering division until the sum of cluster square errors converges to obtain the final clustering result.

[0094] In a specific embodiment, the DBSCAN algorithm is used to identify abnormal data. The DBSCAN algorithm is a density-based noise spatial clustering algorithm. The algorithm reflects the compactness of data distribution through two core parameters: the neighborhood radius ε and the minimum number of sample points MinPts. It can cluster densely accumulated scattered points (i.e., high-density data areas) and regard discrete points as noise points. It is very suitable for abnormal load data identification. The specific steps are as follows:

[0095] For the historical load data set obtained, the neighborhood radius ε and the minimum number of sample points MinPts are set, ε(p) represents the neighborhood with p as the center and ε as the radius, and MinPts represents the minimum number of samples contained in ε(p); a sample point p is randomly selected from the historical load data set. If there are at least MinPts data points in the neighborhood ε(p) of p, the point is recorded as a core object; if the data points contained in the neighborhood ε(p) of p are less than MinPts, and p is not located in other neighborhoods, the point is recorded as a noise point, that is, abnormal data; arbitrarily select unprocessed points and repeat the above noise point judgment operation until all sample points in the historical load data set are processed.

[0096] In general, the neighborhood radius ε and the minimum number of sample points MinPts are determined by the k-dist graph method. The specific steps are as follows:

[0097] Calculate the p for each object in the dataset i The object p that is k closest to the object e The Euclidean distance between them is recorded as the k-dist value of the object, that is,

[0098]

[0099] Among them, n represents the data dimension, that is, the characteristic number (including meteorological factor data, time characteristics, data history load data, such as temperature, air pressure, precipitation, holidays, etc.);

[0100] Arrange the k-dist values ​​of all objects in ascending or descending order and draw the k-dist curve;

[0101] Take the distance corresponding to the first "inflection point" of the k-dist curve as the neighborhood radius ε, and take k+1 as the minimum number of sample points MinPts (to reduce the amount of calculation, the default k=2n-1 is generally used).

[0102] In a specific embodiment, step 3 replaces the abnormal load data with the mean value within the same moment class, and the specific formula is as follows:

[0103]

[0104] in, Indicates the abnormal data correction value; N a represents the total number of samples of the a-th class, represents the measured value of feature n of user f at the jth moment.

[0105] In a specific embodiment, the comprehensive correlation analysis index C M The formula is:

[0106]

[0107] Among them, ρ X,Y is the Pearson correlation coefficient, M IC is the maximum information coefficient value;

[0108] When C M When it is greater than the preset threshold σ, it means that there is a strong correlation between the influencing factor variable X and the load Y. The influencing factor with a strong correlation with the load forecast is selected as the input characteristic variable of the final load combination forecasting model.

[0109] In a specific embodiment, the Pearson correlation coefficient value ρ X,Y The calculation formula is expressed as:

[0110]

[0111] Among them, Y i is the actual observed data sequence of the load, X i are characteristic data sequences that may affect load forecasting, l is the length of the sequence, and ρ X,Y It represents the Pearson correlation coefficient value between variable X and loading Y;

[0112] The maximum information coefficient value M IC The calculation formula is expressed as:

[0113]

[0114] Among them, I(X i ,Y j ) is X i With Y i The mutual information between |X i | and |Y i |respectively the sequence X i With Y i The length of min(|X i |,|Y i |) means taking the sequence X i With Y i Minimum length.

[0115] The embodiment of the present invention uses Informer and GCN to extract and predict features in time and space respectively, and then fuses the single prediction results through the variance-covariance combined prediction model to realize the load prediction for a certain period of time in the future. The Informer-GCN combined prediction model is used as a multivariate load prediction model, which mainly includes three modules: a time feature extraction prediction module, a spatial feature extraction prediction module, and a spatiotemporal feature fusion prediction module. The model structure is as follows Figure 4 shown.

[0116] In a specific embodiment, the temporal feature extraction and prediction module is mainly composed of Informer, which is a deep network model suitable for long time series prediction. It builds a deep network through the attention mechanism to extract the long-term dependency of long time series. Its structure is as follows: Figure 2 As shown. The entire model is divided into two parts: encoder and decoder. The encoder uses a position embedding layer to time encode the input sequence, and then uses multiple neural network modules composed of multi-head probabilistic sparse self-attention and convolutional layers to extract deep features. The load-related variable information input to the decoder is calculated with the deep feature map extracted from the encoder to capture the correlation between features, and finally the estimated load forecast value is obtained through the fully connected layer. In this module, the Informer model selects the user's load data of H time steps as input in the predicted load data set (including multivariate sequences such as unit status data and meteorological data), predicts the load at p moments in the future, and obtains the predicted output result.

[0117] In a specific embodiment, the spatial feature extraction prediction module is mainly composed of GCN, which is a deep learning method for processing non-Euclidean spatial graph data. It aggregates the features of itself and neighboring nodes through convolution operations in the adjacency matrix to generate a new node feature vector. Its structure is as follows Figure 3 In this module, the GCN model takes the weighted user adjacency matrix A and the user feature matrix B as input, predicts the load at the next time p, and obtains the output result

[0118] In a specific embodiment, the weighted user adjacency matrix A is expressed as:

[0119] A={R i1 ,R i2 ,…,R if};

[0120] Among them, R if Represents the spatial correlation between user i and user f, which is calculated using the Pearson correlation coefficient method. The calculation formula is:

[0121]

[0122] Among them, P i is the actual observed data sequence of user i’s load; P f is the actual observed data sequence of user f’s load, l is the sequence length;

[0123] The user feature matrix B is expressed as:

[0124]

[0125] Among them, P i,j represents the load measured by user i at the jth moment.

[0126] In a specific embodiment, the variance-covariance combined prediction model calculates the combined weight coefficients of the Informer model and the GCN model respectively, and outputs the final load forecast value at the next p moments according to the calculated combined weights; the variance-covariance combined prediction model is expressed as:

[0127]

[0128] in Represents the sequence of load forecast values ​​of the future p moments predicted by the Informer model and the GCN model, respectively, t represents the sequence of final load forecast values, w T 、w S They are The weight coefficient of T +w S =1;

[0129] Weight coefficient w T 、w S The method to determine is:

[0130] Based on the forecast load data set, the corresponding forecast results are calculated using the Informer model, GCN model and variance-covariance combined forecasting model. and t ;

[0131] Select the measured load data at time t from the forecast load data set to form a sequence Calculate the error series of the three prediction results respectively:

[0132]

[0133] in, e t They are y t The corresponding error sequence;

[0134] The relationship between the weight coefficient and the error sequence is expressed as:

[0135]

[0136] The error variance of the final load forecast sequence is expressed as:

[0137]

[0138] Among them, Var(e t ) is the error variance of the final load forecast sequence, for covariance of w T Var(e t ) to find the minimum value, we get:

[0139]

[0140] w S =1-w T .

[0141] The embodiment of the present invention also discloses a short-term load forecasting system for multiple users in a substation area based on time-space correlation, comprising:

[0142] Data acquisition module: obtains the historical electricity load data set of multiple users in the substation area;

[0143] Data processing module: Use DBSCAN algorithm to identify abnormal load data in historical load data sets, cluster users through K-means algorithm, replace the characteristic value of abnormal load data with the mean value of the class to which the abnormal load data belongs at the same time, and obtain the optimized load data set;

[0144] Feature optimization module: Use comprehensive correlation analysis indicators to quantitatively analyze the correlation between each load influencing factor and the load in the optimized load data set, select the influencing factors with strong correlation to obtain the predicted load data set as the input of the multivariate load prediction model;

[0145] Load forecasting module: The multivariate load forecasting model uses the Informer model and the GCN model for feature extraction and prediction, and then integrates the prediction results of the Informer model and the GCN model through the variance-covariance combined prediction model to obtain the final load forecasting result.

[0146] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0147] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for short-term load forecasting of multiple users in a substation area based on spatiotemporal correlation, characterized in that: The specific steps are as follows: Obtain the historical electricity load data set of multiple users in the substation area; The DBSCAN algorithm is used to identify abnormal load data in the historical load data set, the users are clustered by the K-means algorithm, and the characteristic value of the abnormal load data is replaced by the mean value within the class to which the abnormal load data belongs at the same moment, so as to obtain an optimized load data set; Using comprehensive correlation analysis indicators, quantitatively analyzing the correlation between each load influencing factor and the load in the optimized load data set, selecting influencing factors with strong correlation to obtain a predicted load data set as an input of a multivariate load prediction model; The multivariate load forecasting model uses the Informer model and the GCN model to perform feature extraction and forecasting, and then fuses the forecast results of the Informer model and the GCN model through a variance-covariance combined forecasting model to obtain a final load forecasting result.

2. The method for short-term load forecasting of multiple users in a substation area based on spatiotemporal correlation according to claim 1 is characterized in that: The specific steps of obtaining the optimized load data set are: Step 1: In the historical load data set, the measured load data of all users in the substation area in the past t time period are intercepted to construct a sample matrix, and the sample matrix is ​​expressed as: Among them, P i,j represents the load measured by user i at the jth moment, P i represents a sample group of the sample matrix P, m represents the total number of users in the area, and k represents the number of sampling points; Step 2: Use the K-means algorithm to divide all sample groups into c user groups a (a=1,2,…,c); Step 3: Replace the characteristic value of the abnormal load data with the load average value in each user group at the same time.

3. The method for short-term load forecasting of multiple users in a substation area based on spatiotemporal correlation according to claim 2 is characterized in that: The step 2 specifically includes: Step 2.1: Randomly select c initial cluster centers P from all sample groups a ; Step 2.2, calculate the Euclidean distance d from each sample group to each of the initial cluster centers, the formula is: Step 2.3: Any sample group P i Divide into the cluster center P that is closest to a In the class represented, traverse all sample groups and complete a division; Step 2.4, calculate the sample mean of each class and calculate the sum of squared errors of the clusters: in, is the sample mean of the ath user group; N a is the total number of samples in the ath class; E is the sum of clustering square errors; Step 2.5, taking the sample mean of each class as the new cluster center, returning to step 2.2 to perform iterative clustering division until the sum of cluster square errors converges to obtain the final clustering result.

4. The method for short-term load forecasting of multiple users in a substation area based on spatiotemporal correlation according to claim 1 is characterized in that: The comprehensive correlation analysis index C M The formula is: Among them, ρ X,Y is the Pearson correlation coefficient, M IC is the maximum information coefficient value; When C M When it is greater than the preset threshold σ, it means that there is a strong correlation between the influencing factor variable X and the load Y. The influencing factor with a strong correlation with the load forecast is selected as the input characteristic variable of the final load combination forecasting model.

5. The method for short-term load forecasting of multiple users in a substation area based on spatiotemporal correlation according to claim 4 is characterized in that: The Pearson correlation coefficient value ρ X,Y The calculation formula is expressed as: Among them, Y i is the actual observed data sequence of the load, X i are characteristic data sequences that may affect load forecasting, l is the length of the sequence, and ρ X,Y It represents the Pearson correlation coefficient value between variable X and loading Y; The maximum information coefficient value M IC The calculation formula is expressed as: Among them, I(X i ,Y j ) is X i With Y i The mutual information between |X i | and |Y i |respectively the sequence X i With Y i The length of min(|X i |,|Y i |) means taking the sequence X i With Y i Minimum length.

6. The method for short-term load forecasting of multiple users in a substation area based on spatiotemporal correlation according to claim 1 is characterized in that: The Informer model selects the user's load data of H time steps in the predicted load data set as input, predicts the load at the next time p, and obtains the predicted output result.

7. The method for short-term load forecasting of multiple users in a substation area based on spatiotemporal correlation according to claim 1 is characterized in that: The GCN model takes the weighted user adjacency matrix A and the user feature matrix B as input, predicts the load at the next time p, and obtains the output result 8. The method for short-term load forecasting of multiple users in a substation area based on spatiotemporal correlation according to claim 7 is characterized in that: The weighted user adjacency matrix A is expressed as: A={R i1 ,R i2 ,…,R if }; Among them, R if Represents the spatial correlation between user i and user f, which is calculated using the Pearson correlation coefficient method. The calculation formula is: Among them, P i is the actual observed data sequence of user i’s load; P f is the actual observed data sequence of user f’s load, l is the sequence length; The user feature matrix B is expressed as: Among them, P i,j Represents the measured load value of user i at a certain time j in history.

9. The method for short-term load forecasting of multiple users in a substation area based on spatiotemporal correlation according to claim 1, characterized in that: The variance-covariance combined prediction model calculates the combined weight coefficients of the Informer model and the GCN model respectively, and outputs the final load forecast value at the next p moments according to the calculated combined weights; the variance-covariance combined prediction model is expressed as: in Respectively represent the sequence of load forecast values ​​predicted by the Informer model and the GCN model at the next p moments, y t represents the sequence of the final load forecast values, w T 、w S They are The weight coefficient of T +w S =1; Weight coefficient w T 、w S The method to determine is: Based on the predicted load data set, the corresponding prediction results are calculated using the Informer model, the GCN model and the variance-covariance combined prediction model. and t ; The measured load data at time t is selected from the predicted load data set to form a sequence Calculate the error series of the three prediction results respectively: in, e t They are y t The corresponding error sequence; The relationship between the weight coefficient and the error sequence is expressed as: The error variance of the final load forecast sequence is expressed as: Among them, Var(e t ) is the error variance of the final load forecast sequence, for covariance of w T Var(e t ) to find the minimum value, we get: In S =1-in T 。 10. A short-term load forecasting system for multiple users in a substation based on spatiotemporal correlation, using a short-term load forecasting method for multiple users in a substation based on spatiotemporal correlation as described in any one of claims 1 to 9, characterized in that: include: Data acquisition module: obtains the historical electricity load data set of multiple users in the substation area; Data processing module: using DBSCAN algorithm to identify abnormal load data in the historical load data set, clustering users through K-means algorithm, replacing the characteristic value of abnormal load data with the mean value of the class to which the abnormal load data belongs at the same moment, and obtaining an optimized load data set; Feature optimization module: using comprehensive correlation analysis indicators, quantitatively analyzing the correlation between each load influencing factor and the load in the optimized load data set, selecting influencing factors with strong correlation to obtain a predicted load data set as the input of the multivariate load prediction model; Load forecasting module: The multivariate load forecasting model uses the Informer model and the GCN model to perform feature extraction and forecasting, and then fuses the forecast results of the Informer model and the GCN model through the variance-covariance combined forecasting model to obtain the final load forecasting result.

Images