Multi-source heterogeneous data fusion method and system oriented to power grid resource business middle station

Abstract

The invention discloses a multi-source heterogeneous data fusion method and system oriented to a power grid resource service intermediate station, and the method comprises the steps: carrying out the preprocessing of the multi-source heterogeneous big data of the power grid resource service intermediate station, and carrying out the feature data extraction of a data source obtained after the preprocessing through a convolution layer; constructing an HFM-SVM (High Frequency Modulation-Support Vector Machine) model, and clustering the extracted feature data; constructing a composite kernel function, and processing the clustered feature data to obtain processed data; classifying the processed data through a sparse auto-encoder, and outputting classified fusion feature data; according to the method provided by the invention, a hybrid fuzzy multi-support vector machine (HFM-SVM) model is constructed, feature fusion of multi-source heterogeneous data is carried out, and the application capability of multi-modal big data in an information system in a power grid resource service is improved.

Description

Technical Field

[0001] The present invention relates to a multi-source heterogeneous data fusion method and system, and in particular to a multi-source heterogeneous data fusion method and system for a power grid resource business middle platform, belonging to the technical field of big data processing. Background Art

[0002] In today's digital age, data has become one of the most valuable assets of enterprises. However, with the explosion of data volume, how to manage and utilize data safely and efficiently, especially data across multiple data centers, has become a major challenge for enterprises. In large enterprise groups, there are usually multiple data centers, which are distributed in different geographical locations and serve different business departments or subsidiaries. These data centers often have their own independent data, and due to business competition, they need to maintain data independence and confidentiality. The monitoring image data of data centers contains privacy data such as relevant equipment nameplates, equipment models, specifications, and configurations. Once these information is leaked, it will expose key business secrets, so it cannot be shared between data centers. Due to the competitive relationship between data centers, sharing these sensitive information will weaken their respective market competitiveness.

[0003] In the power grid field, with the advancement of digital transformation, the use of data is becoming more and more frequent. Digital transformation aims to carry out computational deduction and dynamic presentation of information such as the topological grid, operating status, equipment parameters, and environmental elements of the physical space entity power grid in the digital space through digital technology, based on data elements and digital platforms, so as to comprehensively master the equipment status, comprehensively control the system operation, and comprehensively improve the service quality and efficiency.

[0004] Currently, at the system level, there are problems such as scattered system construction, difficult cross-professional collaborative simulation, and difficult scene connection in the digital calculation and deduction of the power grid. At the same time, the power grid state calculation and deduction faces requirements such as multi-source heterogeneous data fusion, fast analysis and calculation, and high-precision situation deduction. The existing power grid calculation analysis and optimization software technology system has problems such as complex and redundant data organization, low efficiency of storage-computation separation, and difficult in-depth correlation mining, and cannot meet the fast analysis and calculation requirements of the scale of hundreds of millions of equipment nodes and tens of thousands of electrical nodes, making it difficult to perceive key equipment failures in real time, effectively prevent system operation risks, and achieve efficient collaborative optimization of multiple resources.

[0005] In order to effectively integrate these diverse and heterogeneous data of the power grid resource business middle platform, the first thing to solve is the heterogeneous integration of big data, and process the data in multiple levels and all aspects. This not only changes the shortcoming of the single meaning of data, but also ensures the comprehensive accuracy of information in decision-making, which is of great significance for the construction of a new power system and the digital transformation of the power grid. Summary of the Invention

[0006] Objective of the Invention: The objective of the present invention is to provide a multi-source heterogeneous data fusion method and system for a power grid resource business middle platform, which can achieve feature-level data clustering processing of multi-source heterogeneous big data.

[0007] Technical Solution: A multi-source heterogeneous data fusion method for a power grid resource business middle platform according to the present invention includes:

[0008] Preprocessing multi-source heterogeneous big data of the power grid resource business middle platform, and extracting feature data from the data source obtained after preprocessing through a convolutional layer;

[0009] Constructing an HFM-SVM model according to the extracted feature data, and clustering the extracted feature data;

[0010] Constructing a composite kernel function to process the clustered feature data to obtain processed data;

[0011] Classifying the processed data through a sparse autoencoder, and outputting the fused feature data after classification.

[0012] Further, the extracting of the features of the preprocessed data through the convolutional layer includes:

[0013] Representing the convolution kernel in the convolutional layer with a Gaussian function, and the formula is:

[0014]

[0015] where x is the input of the convolutional layer, and σ is the variance of x;

[0016] The formula for extraction through the convolutional layer is:

[0017] output k (i, j) = tanh(Conv k (i, j))

[0018] where output k () is the output of the convolutional layer, tanh is the activation function, the activation value is in the interval [0, 1], k represents the number of convolutional layers, and Conv k (i, j) represents the output value at the position (i, j) after convolutional processing of the input data.

[0019] Further, the constructing of the HFM-SVM model according to the extracted feature data and clustering the extracted feature data includes:

[0020] Selecting d feature data from n groups of data sources as input vectors: x dn(n = 1, 2, ..., N, d = 1, 2, ..., D), where N and D represent the number of data sources and the amount of feature data corresponding to each data source, respectively;

[0021] Using the fuzzy c-means (FCM) algorithm, the input vector x dn is partitioned into C clusters, and the feature vectors of the cluster centers are denoted as c j , 1 ≤ j ≤ C, and the corresponding fuzzy partition matrix u satisfies the condition:

[0022]

[0023] The constructed HFM-SVM model aims to minimize the objective function value of the following formula:

[0024]

[0025] where f i is the feature vector of the i-th data point, ||*|| represents the Euclidean distance, m is the fuzzy coefficient, represents the membership degree of f i belonging to cluster j; c j and u ij The updated formulas are as follows:

[0026]

[0027] Furthermore, the construction of the composite kernel function for processing the clustered feature data includes:

[0028] Let the composite kernel function be:

[0029]

[0030] where K(u j , u k ) represents the kernel function value between vectors u j and u k , u j and u k are the feature vectors of the input vector matrix and the center vector matrix respectively, γ is the kernel parameter, ||*|| represents the Euclidean distance, C is the total number of clusters, and T represents the transpose.

[0031] Furthermore, the classification of the processed heterogeneous data through the sparse autoencoder to output the classified fusion feature data specifically includes:

[0032] Using the stacked autoencoder and the reconstructed data to create a sparse autoencoder, classifying the processed data, determining the centroid and category of the processed data, and obtaining the classification result of the data after several iterations, and outputting the classified fusion feature data.

[0033] Based on the same inventive concept, the present invention also provides a multi-source heterogeneous data fusion system for a power grid resource service middle platform, including:

[0034] A feature extraction module, configured to preprocess the multi-source heterogeneous big data of the power grid resource service middle platform, and extract feature data from the data source obtained after preprocessing through a convolutional layer;

[0035] A clustering module, configured to construct an HFM-SVM model according to the extracted feature data, and cluster the extracted feature data;

[0036] A processing module, configured to construct a composite kernel function to process the clustered feature data to obtain processed data;

[0037] A fusion module, configured to classify the processed data through a sparse autoencoder and output the fused feature data after classification.

[0038] Further, the extraction of the features of the preprocessed data by the feature extraction module through the convolutional layer includes:

[0039] Represent the convolution kernel in the convolutional layer with a Gaussian function, and the formula is:

[0040]

[0041] where x is the input of the convolutional layer, and σ is the variance of x;

[0042] The formula for extraction through the convolutional layer is:

[0043] output k (i, j) = tanh(Conv k (i, j))

[0044] where output k () is the output of the convolutional layer, tanh is the activation function, the activation value is in the interval [0, 1], k represents the number of convolutional layers, and Conv k (i, j) represents the output value at the position (i, j) after convolutional processing of the input data.

[0045] Further, the clustering module includes:

[0046] Select d feature data from n groups of data sources as input vectors: x dn (n = 1, 2,..., N, d = 1, 2,..., D), where N and D respectively represent the number of data sources and the amount of feature data corresponding to each data source;

[0047] Use the fuzzy clustering FCM algorithm to process the input vector x dnDivided into C clusters, the feature vectors of the cluster centers are represented as c j , 1 ≤ j ≤ C, and the corresponding fuzzy partition matrix u satisfies the condition:

[0048]

[0049] The constructed HFM-SVM model aims to minimize the objective function value of the following formula:

[0050]

[0051] where f i is the feature vector of the i-th data point, ||*|| represents the Euclidean distance, m is the fuzzy coefficient, represents the membership degree that f i belongs to cluster j; c j and u ij The updated formulas are as follows:

[0052]

[0053] Furthermore, the processing module includes:

[0054] Let the composite kernel function be:

[0055]

[0056] where K(u j , u k ) represents the kernel function value between vectors u j and u k , u j and u k are the feature vectors of the input vector matrix and the center vector matrix respectively, γ is the kernel parameter, ||*|| represents the Euclidean distance, C is the total number of clusters, and T represents the transpose.

[0057] Furthermore, the fusion module includes:

[0058] Create a sparse autoencoder using a stacked autoencoder and reconstructed data, classify the processed data, determine the centroid and category of the processed data, and obtain the classification result of the data after several iterations, and output the fused feature data after classification.

[0059] Based on the same inventive concept, the present invention also provides a computing device, including: one or more processors, one or more memories, and one or more programs, the programs are stored in the memory and configured to be executed by the processor, and when the programs are loaded into the processor, the steps of the multi-source heterogeneous data fusion method for the power grid resource business middle platform according to any one of the above are implemented.

[0060] Based on the same inventive concept, the present invention also provides a storage medium storing a computer program including program instructions, which when executed by a processor cause the processor to execute the steps of the multi-source heterogeneous data fusion method for the power grid resource service middle platform according to any one of the above.

[0061] Advantageous effects: Compared with the prior art, the present invention combines a multi-support vector machine and a convolutional neural network to achieve clustering processing of feature-level data of multi-source heterogeneous big data, providing support for data processing and higher-level system decision-making of the power grid resource service middle platform; through the fusion processing of multi-source heterogeneous big data, intelligent management of power grid big data is effectively achieved; through the heterogeneous data fusion framework combined with different types of data features, it can be extended and applied to more fields, with strong generalization performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Figure 1 It is a flowchart of the method according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0063] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application.

[0064] In the technical solution of this embodiment, there are mainly several parts including data feature extraction, data clustering, data processing and fusion. As shown in the accompanying Figure 1 drawings, the multi-source heterogeneous data fusion method for the power grid resource service middle platform in this embodiment includes:

[0065] Step 1: Preprocess the multi-source heterogeneous big data of the power grid resource service middle platform, and extract feature data from the data source obtained after preprocessing through a convolutional layer;

[0066] Step 2: Construct an HFM-SVM model according to the extracted feature data, and cluster the extracted feature data;

[0067] Step 3: Construct a composite kernel function to process the clustered feature data.

[0068] Step 4: Classify the heterogeneous data processed in Step 3 through a sparse autoencoder, and output the feature data after fusion classification.

[0069] Further, the specific content of Step 1 includes:

[0070] The convolution kernel in the convolutional layer is a Gaussian function, and the specific formula is:

[0071]

[0072] Among them, x represents the input data in the convolutional layer, and the input data for the convolutional layer operation is the preprocessed data of the power grid resource business middle platform. σ is the variance of x.

[0073] The formula for processing data through the convolutional layer is:

[0074] output k (i, j) = tanh(Conv k (i, j))

[0075] Among them, output k () is the feature data output by the convolutional layer. tanh is an activation function, and its activation value is in the interval [0, 1]. k represents the number of convolutional layers. Conv k (i, j) represents the output value at the position (i, j) after convolutional processing of the input data.

[0076] The activation function introduces non-linear characteristics, enabling the constructed algorithm model to learn and represent complex non-linear relationships and improving the processing ability for heterogeneous data

[0077] Furthermore, step 2 specifically includes:

[0078] Step 2.1: Extract features from multiple heterogeneous data sources to train the model, that is, select d input vectors x from n groups of data sources dn (n = 1, 2,..., N, d = 1, 2,..., D), where N and D respectively represent the number of heterogeneous data sources and the amount of feature data corresponding to each data source;

[0079] Step 2.2: Use the fuzzy c-means (FCM) algorithm to divide the input vector x dn into C clusters, and the feature vector of the cluster center is represented as c j , 1 ≤ j ≤ C, and the corresponding fuzzy partition matrix u should satisfy the condition:

[0080]

[0081] Step 2.3: The core objective of the constructed HFM-SVM model is to minimize the following objective function value, that is:

[0082]

[0083] Among them, f i is the feature vector of the i-th data point, ||*|| represents the Euclidean distance, the fuzzy coefficient is represented as m, represents f iThe membership degree belonging to cluster j, usually m > 1, which determines the fuzziness of the membership degree value. When m is larger, the difference in membership degree values is more obvious, and the clustering result is more fuzzy. c j and u ij The updated formula is as follows:

[0084]

[0085] Furthermore, the specific steps of step 3 include:

[0086] The kernel function is the key for the support vector machine to transform from a low-dimensional space to a high-dimensional space, which can make the data in the original feature space linearly separable in the high-dimensional space, effectively avoiding the insoluble problem caused by over-dimensionality. Therefore, it is necessary to determine a suitable kernel function and kernel parameter to solve the specific problem of heterogeneous data fusion. The composite kernel function selected in this patent is expressed as:

[0087]

[0088] Among them, K(u j , u k ) represents the kernel function value between the input vectors u j , u k . u j , u k are the eigenvectors of the input vector matrix and the center vector matrix respectively, and γ is the kernel parameter. The kernel parameter is very important and determines the bias and generalization ability of the model. If the kernel parameter value is too large or too small, it will cause problems such as the constructed model being overly sensitive or overly insensitive to noise samples. In this patent, the kernel parameter value is solved by the trial-and-error method. First, set the initial value of the kernel parameter, enter the model for model training and testing, then modify and optimize the kernel parameter value according to the model prediction accuracy, and transfer the adjusted parameter into the model for repeated training and testing to ensure that the best fusion effect can be obtained for data fusion.

[0089] Specifically, in the experiment, the average deviation value, the maximum deviation value, and the prediction accuracy of the model prediction were calculated when the kernel parameter γ was taken. The experimental results are shown in Table 1:

[0090] Table 1 Experimental result parameter table

[0091] Nuclear parameter γ Average deviation value Maximum deviation value Prediction accuracy 1 1.13 6 42.93 3 1.56 6 42.93 5 1.28 8 44.98 10 1.41 8 38.6 15 1.36 8 37.8 20 1.31 8 45.04 30 1.25 8 45.53 50 1.25 8 48.52

[0092] It can be seen from the data in the table that when the kernel parameter γ = 30, the average deviation value is the smallest and the prediction accuracy reaches the maximum. Therefore, the final value of the kernel parameter γ is set to 30.

[0093] Furthermore, the specific steps of step 4 include:

[0094] The stacked autoencoder and reconstructed data are used to create a sparse autoencoder, the output data in step 3 is processed, the centroid and category of the input data are determined, and the classification results of the data are obtained after multiple iterations.

[0095] Since big data is highly specialized and there is no obvious division of data, in order to avoid excessive interference in the data fusion process, it is necessary to build data fusion when the sparse model is sparse to improve the generalization ability and interpretability of the model. Therefore, this patent adopts a sparse autoencoder based on clustering analysis to fuse heterogeneous data so that the output data can maintain the original characteristics as much as possible.

[0096] Based on the same inventive concept, this embodiment also provides a multi-source heterogeneous data fusion system for a power grid resource service center, including:

[0097] The feature extraction module is used to pre-process the multi-source heterogeneous big data of the power grid resource business platform, and extract feature data from the data source obtained after pre-processing through the convolution layer;

[0098] A clustering module is used to construct an HFM-SVM model based on the extracted feature data and cluster the extracted feature data;

[0099] A processing module is used to construct a composite kernel function to process the clustered feature data to obtain processed data;

[0100] The fusion module is used to classify the processed data through the sparse autoencoder and output the classified fusion feature data.

[0101] Furthermore, the feature extraction module extracts features of the preprocessed data through a convolutional layer, including:

[0102] The convolution kernel in the convolution layer is represented by a Gaussian function, and the formula is:

[0103]

[0104] Where x is the input of the convolutional layer and σ is the variance of x;

[0105] The formula extracted by the convolutional layer is:

[0106] output k (i, j) = tanh(Conv k (i, j))

[0107] Among them, output k () is the output of the convolutional layer, tanh is the activation function, the activation value is in the range of [0, 1], k represents the number of convolutional layers, Conv k(i, j) represents the output value at position (i, j) after convolutional processing of the input data.

[0108] Furthermore, the clustering module includes:

[0109] Select d feature data from n groups of data sources as the input vector: x dn (n = 1, 2,..., N, d = 1, 2,..., D), where N and D respectively represent the number of data sources and the amount of feature data corresponding to each data source;

[0110] Use the fuzzy c-means (FCM) clustering algorithm to divide the input vector x dn into C clusters, and the feature vector of the cluster center is represented as c j , 1 ≤ j ≤ C, and the corresponding fuzzy partition matrix u satisfies the condition:

[0111]

[0112] The constructed HFM-SVM model aims to minimize the objective function value of the following formula:

[0113]

[0114] where f i is the feature vector of the i-th data point, ||*|| represents the Euclidean distance, m is the fuzzy coefficient, represents the membership degree of f i belonging to cluster j; c j and u ij The updated formulas are as follows:

[0115]

[0116] Furthermore, the processing module includes:

[0117] Let the composite kernel function be:

[0118]

[0119] where K(u j , u k ) represents the kernel function value between vectors u j and u k , u j and u k are the feature vectors of the input vector matrix and the center vector matrix respectively, γ is the kernel parameter, ||*|| represents the Euclidean distance, C is the total number of clusters, and T represents the transpose.

[0120] Furthermore, the fusion module includes:

[0121] Create a sparse autoencoder by using a stacked autoencoder and reconstructed data, classify the processed data, determine the centroid and category of the processed data, and obtain the classification result of the data after several iterations, and output the fused feature data after classification.

[0122] Based on the same inventive concept, this embodiment further provides a computing device, including: one or more processors, one or more memories, and one or more programs, the programs are stored in the memory and configured to be executed by the processors, and when the programs are loaded into the processors, the steps of the multi-source heterogeneous data fusion method for the power grid resource business middle platform according to any one of the above are implemented.

[0123] Based on the same inventive concept, this embodiment further provides a storage medium, the storage medium stores a computer program, the computer program includes program instructions, and when the program instructions are executed by a processor, the processor is caused to execute the steps of the multi-source heterogeneous data fusion method for the power grid resource business middle platform according to any one of the above.

[0124] This embodiment constructs a hybrid fuzzy multi-support vector machine (HFM-SVM) model to perform feature fusion of multi-source heterogeneous data, improving the application ability of multi-modal big data in the information system of the power grid resource business middle platform.

Claims

1. A multi-source heterogeneous data fusion method for power grid resource business center, characterized in that: include: Preprocess the multi-source heterogeneous big data of the power grid resource business platform, and extract feature data from the preprocessed data source through the convolution layer; According to the extracted feature data, the HFM-SVM model is constructed to cluster the extracted feature data; Construct a composite kernel function to process the clustered feature data to obtain processed data; The processed data is classified through the sparse autoencoder, and the classified fusion feature data is output.

2. The multi-source heterogeneous data fusion method for power grid resource business center according to claim 1 is characterized in that: The extracting the features of the preprocessed data through the convolution layer includes: The convolution kernel in the convolution layer is represented by a Gaussian function, and the formula is: Where x is the input of the convolutional layer and σ is the variance of x; The formula extracted by the convolutional layer is: output k (i,j)=tanh(Conv k (i,j)) Among them, output k () is the output of the convolutional layer, tanh is the activation function, the activation value is in the range of [0,1], k represents the number of convolutional layers, Conv k (i, j) represents the output value at position (i, j) after convolution processing of the input data.

3. The multi-source heterogeneous data fusion method for power grid resource business center according to claim 1 is characterized in that: The HFM-SVM model is constructed based on the extracted feature data to cluster the extracted feature data, including: Select d feature data from n groups of data sources as input vector: x dn (n=1,2,…,N,d=1,2,…,D), N and D represent the number of data sources and the amount of feature data corresponding to each data source respectively; Using the fuzzy clustering FCM algorithm, the input vector x dn Divided into C clusters, the characteristic vector of the cluster center is represented by c j ,1≤j≤C, the corresponding fuzzy partitioning matrix u satisfies the condition: The constructed HFM-SVM model aims to minimize the following objective function value: Among them, f i is the feature vector of the i-th data point, ||*|| represents the Euclidean distance, m is the fuzzy coefficient, represents f i The membership degree of cluster j; c j and u ij The updated formula is as follows:

4. The multi-source heterogeneous data fusion method for power grid resource business center according to claim 1 is characterized in that: The step of constructing a composite kernel function to process the clustered feature data includes: Assume the composite kernel function is: Among them, K(u j ,u k ) represents the vector u j and u k The kernel function value between j and u k are the eigenvectors of the input vector matrix and the center vector matrix respectively, γ is the kernel parameter, ||*|| represents the Euclidean distance, C is the total number of clusters, and T represents the transpose.

5. The multi-source heterogeneous data fusion method for power grid resource business center according to claim 1 is characterized in that: The sparse autoencoder is used to classify the processed heterogeneous data and output the classified fusion feature data, which specifically includes: A sparse autoencoder is created using stacked autoencoders and reconstructed data to classify the processed data, determine the centroid and category of the processed data, and obtain the classification results of the data after several iterations, and output the classified fusion feature data.

6. A multi-source heterogeneous data fusion system for power grid resource business middle station, characterized in that: include: The feature extraction module is used to pre-process the multi-source heterogeneous big data of the power grid resource business platform, and extract feature data from the data source obtained after pre-processing through the convolution layer; A clustering module is used to construct an HFM-SVM model based on the extracted feature data and cluster the extracted feature data; A processing module is used to construct a composite kernel function to process the clustered feature data to obtain processed data; The fusion module is used to classify the processed data through the sparse autoencoder and output the classified fusion feature data.

7. The multi-source heterogeneous data fusion system for power grid resource business middle station according to claim 6 is characterized in that: The feature extraction module extracts features of the preprocessed data through a convolutional layer, including: The convolution kernel in the convolution layer is represented by a Gaussian function, and the formula is: Where x is the input of the convolutional layer and σ is the variance of x; The formula extracted by the convolutional layer is: output k (i,j)=tanh(Conv k (i,j)) Among them, output k () is the output of the convolutional layer, tanh is the activation function, the activation value is in the range of [0,1], k represents the number of convolutional layers, Conv k (i, j) represents the output value at position (i, j) after convolution processing of the input data.

8. The multi-source heterogeneous data fusion system for power grid resource business middle station according to claim 6 is characterized in that: The clustering module comprises: Select d feature data from n groups of data sources as input vector: x dn (n=1,2,…,N,d=1,2,…,D), N and D represent the number of data sources and the amount of feature data corresponding to each data source respectively; Using the fuzzy clustering FCM algorithm, the input vector x dn Divided into C clusters, the characteristic vector of the cluster center is represented by c j ,1≤j≤C, the corresponding fuzzy partitioning matrix u satisfies the condition: The constructed HFM-SVM model aims to minimize the following objective function value: Among them, f i is the feature vector of the i-th data point, ||*|| represents the Euclidean distance, m is the fuzzy coefficient, represents f i The membership degree of cluster j; c j and u ij The updated formula is as follows:

9. The multi-source heterogeneous data fusion system for power grid resource business middle station according to claim 6 is characterized in that: The processing module comprises: Assume the composite kernel function is: Among them, K(u j ,u k ) represents the vector u j and u k The kernel function value between j and u k are the eigenvectors of the input vector matrix and the center vector matrix respectively, γ is the kernel parameter, ||*|| represents the Euclidean distance, C is the total number of clusters, and T represents the transpose.

10. The multi-source heterogeneous data fusion system for power grid resource business middle station according to claim 6 is characterized in that: The fusion module comprises: A sparse autoencoder is created using stacked autoencoders and reconstructed data to classify the processed data, determine the centroid and category of the processed data, and obtain the classification results of the data after several iterations, and output the classified fusion feature data.

11. A computing device, characterized in that: include: One or more processors, one or more memories, and one or more programs, wherein the programs are stored in the memories and configured to be executed by the processors, and when the programs are loaded into the processors, the steps of the multi-source heterogeneous data fusion method for the power grid resource business center according to any one of claims 1 to 5 are implemented.

12. A storage medium, characterized in that: The storage medium stores a computer program, which includes program instructions. When the program instructions are executed by a processor, the processor executes the steps of the multi-source heterogeneous data fusion method for a power grid resource business center according to any one of claims 1 to 5.

Images