Virtual power plant user sensitive data identification method and system, medium and terminal

Abstract

The invention discloses a virtual power plant user sensitive data identification method and system, a medium and a terminal. The method comprises the steps of obtaining a to-be-identified user data graph of a virtual power plant; inputting the to-be-recognized user data graph into a sensitive data recognition model, so that the model performs recognition processing on the to-be-recognized user data graph and outputs a sensitive level of the to-be-recognized user data graph; wherein the sensitive data identification model is formed by introducing a graph attention mechanism on the basis of a graph convolutional layer of a graph convolutional neural network model; and the graph convolution layer introducing the graph attention mechanism is used for distributing weights for the nodes of the user data graph to be recognized input into the sensitive data recognition model based on the attention mechanism and in combination with the attention coefficients of the adjacent nodes so as to extract node features in the user data graph to be recognized. According to the method, the graph attention mechanism is introduced on the basis of the graph convolutional neural network model, so that the generalization ability of the graph neural network model is improved, and the accuracy of user sensitive data recognition is improved.

Description

Technical Field

[0001] The present invention relates to the field of power grid data security technology, and in particular to a method, system, medium and terminal for identifying sensitive data of virtual power plant users. Background Art

[0002] The energy transformation strategy has brought different challenges and opportunities to the development of all walks of life. For the power industry, the challenge lies in how to achieve the goal of promoting the consumption of clean energy and optimizing the allocation of existing resources. The opportunity lies in how to seize the opportunity and stimulate the potential of interactive consumption of power energy through deepening the reform of the power market. Many experts and scholars have given full play to the spirit of "pioneers". Practice has proved that the virtual power plant model is a favorable weapon for the integration and development of challenges and opportunities. The security protection of user data of complex virtual power plants has become an important foundation for the State Grid's data security protection work. With the promulgation of various data protection laws, data identification requirements have been clearly put forward. In recent years, power grid companies have actively made a series of preliminary explorations around data identification, and have initially formed identification methods and practical operation foundations.

[0003] The existing schemes for identifying sensitive data in virtual power plants mainly generate a graph network to be identified based on the relationship between keywords in the user data to be identified, and then input the graph network to be identified into a graph convolutional neural network to obtain the minimum connected dominating set corresponding to the graph network to be identified, and then generate a simplified graph of the graph network to be identified based on the minimum connected dominating set, and finally calculate the matching degree between the simplified graph of the graph network to be identified and the simplified graph corresponding to the standard sensitive level data to obtain the sensitive identification result of the user data to be identified. However, when identifying sensitive data in virtual power plants, the graph network established by the above scheme has weak generalization ability and cannot meet the accuracy requirements of sensitive data identification in virtual power plants. Summary of the invention

[0004] The embodiments of the present invention provide a method, system, medium and terminal for identifying sensitive data of virtual power plant users, which introduces a graph attention mechanism based on the graph convolution layer of the graph convolutional neural network model to improve the generalization ability of the graph neural network model, thereby improving the accuracy of identifying user sensitive data.

[0005] In order to solve the above technical problems, an embodiment of the present invention provides a method for identifying sensitive data of virtual power plant users, comprising:

[0006] Obtaining a user data graph to be identified in the virtual power plant;

[0007] Inputting the user data graph to be identified into a pre-built sensitive data identification model, so that the sensitive data identification model performs identification processing on the user data graph to be identified and outputs the sensitivity level of the user data graph to be identified as a sensitive data identification result;

[0008] Among them, the sensitive data identification model is constructed by introducing a graph attention mechanism on the basis of the graph convolution layer of the graph convolutional neural network model; the graph convolution layer introducing the graph attention mechanism is used to assign weights to the nodes of the user data graph to be identified that is input into the sensitive data identification model based on the attention mechanism and combined with the attention coefficients of adjacent nodes, so as to extract the node features in the user data graph to be identified.

[0009] In the implementation of the present invention, a graph attention mechanism is introduced on the basis of the graph convolution layer of the graph convolution neural network model to form a sensitive data recognition model, and a graph convolution layer of the graph attention mechanism is introduced to allocate weights to each node of the user data graph to be identified input into the sensitive data recognition model based on the attention mechanism and in combination with the attention coefficients of the adjacent nodes, so as to extract the node features in the user data graph to be identified. By the above means, the graph attention mechanism is introduced into the graph convolution neural network model, which can change the defect that the graph convolution neural network model can only allocate the same weight to adjacent intersection points of the same order, so that the graph neural network model has a better generalization ability and improves the accuracy of user sensitive data identification. After the optimization of the graph convolution layer of the graph convolution neural network model is completed to obtain the sensitive data recognition model, the user data graph to be identified of the virtual power plant is input into the sensitive data recognition model, so that the sensitive data recognition model performs recognition processing on the user data graph to be identified and outputs the sensitivity level of the user data graph to be identified as the sensitive data recognition result, so that the operation and maintenance management personnel can timely and accurately grasp the data security of the virtual power plant and provide data support for the optimization measures of the virtual power plant. In addition, through the graph convolution layer, the model can utilize the node and edge information in the graph structure to capture the contextual relationships between nodes, enabling the model to more accurately identify and predict complex relationships and patterns, especially in dynamically changing scenarios.

[0010] As a preferred solution, the extraction of node features in the to-be-identified user data graph is specifically as follows:

[0011] Using the graph attention mechanism, all nodes of the user data graph to be identified that is input into the sensitive data identification model are calculated to obtain the correlation between the nodes;

[0012] Traversing each node of the to-be-identified user data graph, performing softmax normalization processing on the correlation between the current traversal node and the adjacent nodes of the current traversal node, and obtaining the attention coefficient corresponding to the current traversal node;

[0013] According to the preset attention mechanism algorithm, based on the attention coefficient corresponding to the current traversal node and the output vector matrix of the adjacent nodes of the current traversal node, the output vector of the current traversal node is calculated as the node feature corresponding to the current traversal node.

[0014] By implementing the preferred solution of the embodiment of the present invention, the model can adaptively assign different attention weights to the neighboring nodes of each node through the graph attention mechanism, and realize the dynamic weight allocation of nodes, so that the model can pay more attention to the neighboring nodes that have a greater impact on the currently traversed node, thereby improving the accuracy of feature extraction. In addition, through softmax normalization processing, the attention coefficient can more clearly reflect the correlation between nodes.

[0015] As a preferred solution, the output vector of the current traversal node is calculated according to the preset attention mechanism algorithm based on the attention coefficient corresponding to the current traversal node and the output vector matrix of the adjacent nodes of the current traversal node, as the node feature corresponding to the current traversal node, specifically:

[0016] h′ i =σ(∑a ij W j );

[0017] In the formula, h′ i represents the output vector of the i-th node; σ represents the activation function; a ij is the attention coefficient of the i-th node and its neighbor node j; W is the weight parameter matrix; h j is the output vector matrix of node j.

[0018] By implementing the preferred solution of the embodiment of the present invention, through the combination of activation function and weight parameter matrix, the model can better capture the complex relationship between nodes, thereby improving the overall performance of the model, especially on high-dimensional or complex data sets. In addition, in traditional graph neural networks, all adjacent nodes contribute equally to the central node, and after introducing the attention mechanism, the model can automatically filter out noise nodes, reduce the impact of noise on the model prediction results, and improve the accuracy and reliability of the prediction.

[0019] As a preferred solution, the expression of the graph attention mechanism is specifically:

[0020] Attention(Q, S) = Σ <Q,S i > ·S i ;

[0021] In the formula, S represents the user data graph to be identified in the virtual power plant, Q represents the sensitive data graph of known users in the virtual power plant, and S i Represents the i-th information in the user data graph to be identified of the virtual power plant.

[0022] By implementing the preferred scheme of the embodiment of the present invention, by combining the data graph of the user to be identified and the sensitive data graph of the known user, the model can better integrate and utilize known information to identify and predict unknown data. This integration capability helps to improve the accuracy and efficiency of sensitive data identification.

[0023] As a preferred solution, the graph convolutional neural network model includes: an input layer, a graph convolution layer, a fully connected layer, a batch normalization layer, a spatiotemporal convolution layer, an average pooling layer, a maximum pooling layer and an output layer;

[0024] Among them, the spatiotemporal convolution layer is used to perform convolution operations on the spatiotemporal data of each node; the spatiotemporal data of each node is spatiotemporal feature data formed by introducing time dimension and space dimension into the node feature of each node.

[0025] The preferred solution of implementing the embodiment of the present invention is to introduce the spatiotemporal convolution layer, so that the model can process data in both the time dimension and the space dimension at the same time, which means that the model can not only capture the spatial relationship between nodes (such as geographic location), but also capture the behavior pattern of nodes changing over time, thereby realizing multi-dimensional feature extraction, so as to more comprehensively understand and analyze the data. In addition, the fully connected layer can comprehensively process the features extracted by the previous layers to generate the final prediction results. This comprehensive processing capability enables the model to better adapt to different task requirements, such as classification, regression, anomaly detection, etc. The batch normalization layer can accelerate the training process and reduce the risk of overfitting of the model, while the average pooling layer and the maximum pooling layer can further extract important features, reduce the dimension of the data, and reduce the computational complexity.

[0026] As a preferred solution, the method of obtaining the to-be-identified user data map of the virtual power plant is specifically as follows:

[0027] Collect virtual power plant user data in real time within a preset time period;

[0028] Keywords are extracted from the user data of the virtual power plant, and the extracted keywords are used as nodes. The contextual relationships between the keywords are used as edges between the nodes. Based on all nodes and the edges between the nodes, a user data graph to be identified of the virtual power plant is constructed.

[0029] A preferred scheme for implementing the embodiment of the present invention collects virtual power plant user data in real time and extracts keywords, uses the extracted keywords as nodes, uses the contextual relationships between the keywords as edges between the nodes, and constructs a user data graph to be identified for the virtual power plant based on all nodes and the edges between the nodes. This can integrate user data scattered in different systems into a unified graph structure, so as to intuitively display the correlation between virtual power plant user data, thereby enabling more rapid and accurate identification of abnormal user data.

[0030] In order to solve the same technical problem, an embodiment of the present invention further provides a system for identifying sensitive data of virtual power plant users, including:

[0031] A data acquisition module, used to acquire a data graph of users to be identified in a virtual power plant;

[0032] A sensitive data identification module is used to input the user data graph to be identified into a pre-built sensitive data identification model, so that the sensitive data identification model identifies and processes the user data graph to be identified and outputs the sensitivity level of the user data graph to be identified as a sensitive data identification result; wherein, the sensitive data identification model is constructed by introducing a graph attention mechanism on the basis of a graph convolutional layer of a graph convolutional neural network model; the graph convolutional layer that introduces the graph attention mechanism is used to assign weights to the nodes of the user data graph to be identified that is input into the sensitive data identification model based on the attention mechanism and in combination with the attention coefficients of adjacent nodes, so as to extract the node features in the user data graph to be identified.

[0033] As a preferred solution, the data acquisition module specifically includes:

[0034] A data collection unit, used to collect virtual power plant user data of the virtual power plant in a preset time period in real time;

[0035] A data graph construction unit is used to extract keywords from the user data of the virtual power plant, and use the extracted keywords as nodes, and the contextual relationship between each keyword as the edge between each node, and construct the user data graph to be identified of the virtual power plant based on all nodes and the edges between each node.

[0036] In order to solve the same technical problem, the present invention also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute the method for identifying sensitive data of virtual power plant users.

[0037] In order to solve the same technical problem, the present invention also provides a terminal, including a processor, a memory and a computer program stored in the memory; wherein the computer program can be executed by the processor to implement the method for identifying sensitive data of virtual power plant users. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 : A flowchart of a method for identifying sensitive data of a virtual power plant user provided in Embodiment 1 of the present invention;

[0039] Figure 2: A structural diagram of a virtual power plant user sensitive data identification system provided in Example 1 of the present invention. DETAILED DESCRIPTION

[0040] 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.

[0041] Embodiment one:

[0042] Please refer to Figure 1 , is a method for identifying sensitive data of virtual power plant users provided by an embodiment of the present invention, the method comprises steps S1 to S2, each step is specifically as follows:

[0043] Step S1, obtaining a data map of users to be identified in the virtual power plant.

[0044] As a preferred solution, step S1 includes step S11 to step S12, and each step is specifically as follows:

[0045] Step S11, collecting virtual power plant user data of the virtual power plant within a preset time period in real time.

[0046] In this embodiment, the power data collection layer is used to extract data such as the operation status of user equipment, transmission quota, energy type, power price, source address and receiving time of the virtual power plant within a period of time as the virtual power plant user data within a preset time period. Among them, the power data collection layer is the physical area and basic module of the virtual power plant power data management platform.

[0047] Step S12, extract keywords from the virtual power plant user data, and use the extracted keywords as nodes, and the contextual relationship between each keyword as the edge between each node, and build a user data graph to be identified for the virtual power plant based on all nodes and the edges between each node.

[0048] It should be noted that a graph is a collection of vertices and edges, or a graph is a collection of nodes and the relationships between these nodes. Nodes generally point to research entities, while edges point to the connection relationship between research entities, which can be entity connections.

[0049] Step S2, inputting the user data graph to be identified into a pre-built sensitive data identification model, so that the sensitive data identification model performs identification processing on the user data graph to be identified and outputs the sensitivity level of the user data graph to be identified as a sensitive data identification result.

[0050] In this embodiment, the sensitivity levels include four sensitivity levels: severe, high, medium, and low.

[0051] As an example, the process of "sensitive data identification model identifying and processing the user data graph to be identified" is as follows: based on the user data graph to be identified, the adjacent node feature information corresponding to the adjacent nodes of the target node is generated, and then the target node's own feature information is updated through the adjacent node feature information, so that the coal seam of the graph neural network model radiates its own information to the first-order adjacent nodes after information aggregation and transformation, thereby improving the graph convolutional neural network model's ability to diffuse local information to the whole.

[0052] It should be noted that the graph convolutional neural network model mainly uses the graph Fourier transform to achieve convolution. It uses the graph Laplacian matrix to derive the Laplacian operator in the frequency domain, and then derives the formula for graph convolution by analogy with the convolution in the Euclidean space in the frequency domain. Compared with spectral domain convolution, spatial domain convolution is easier to understand in terms of design concept. Based on the principle of spatial correlation, spatial domain convolution uses the feature information of the target node itself and the feature information of the adjacent nodes to update the feature information of the target node, and completes the "classification" operation by aggregating the information of the surrounding nodes to itself. In the specific implementation process, the graph convolution layer and the graph pooling module can be interlaced to ultimately realize the extraction of high-dimensional features of the graph data, and the initially input graph is processed into a high-level substructure through graph convolution, that is, the representation of each level of the graph or the extraction of feature information at each level, and the extracted information is finally used to output the classification result.

[0053] Among them, the sensitive data recognition model is constructed by introducing a graph attention mechanism on the basis of the graph convolution layer of the graph convolutional neural network model to improve the adaptability of the graph convolutional neural network model; the graph convolution layer introducing the graph attention mechanism is used to assign weights to the nodes of the user data graph to be identified that is input into the sensitive data recognition model based on the attention mechanism and combined with the attention coefficients of adjacent nodes to extract the node features in the user data graph to be identified.

[0054] It should be noted that the graph attention mechanism originates from the field of computer vision and is inspired by the way human cognitive nerves process things. When humans process information, they will selectively focus on the more important parts and ignore the unimportant parts to reduce the work of information processing, thereby improving the efficiency of human information processing. When this processing mechanism is applied to computers, it is necessary to filter out local useful information from a large amount of information, and assign weights to the information to be processed, giving high weights to the parts you want to focus on and low weights to the useless parts.

[0055] As a preferred solution, the above process of “extracting node features in the user data graph to be identified” includes steps S21 to S23, and each step is specifically as follows:

[0056] Step S21, using the graph attention mechanism, calculate all nodes of the user data graph to be identified that is input into the sensitive data recognition model to obtain the correlation between each node.

[0057] As a preferred solution, the mathematical expression of the “graph attention mechanism” mentioned in step S21 is specifically shown in formula (1).

[0058] Attention(Q, S) = ∑ <Q,S i > ·S i ; (1)

[0059] In the formula, S represents the user data graph to be identified in the virtual power plant, Q represents the sensitive data graph of known users in the virtual power plant, and S i Represents the i-th information in the user data graph to be identified of the virtual power plant.

[0060] Step S22, traverse each node of the user data graph to be identified, perform softmax normalization processing on the correlation between the current traversal node and the adjacent nodes of the current traversal node, and obtain the attention coefficient corresponding to the current traversal node.

[0061] Step S23, according to the preset attention mechanism algorithm, based on the attention coefficient corresponding to the current traversal node and the output vector matrix of the adjacent nodes of the current traversal node, the output vector of the current traversal node is calculated as the node feature corresponding to the current traversal node.

[0062] As a preferred solution, the “attention mechanism algorithm” mentioned in step S23, please refer to formula (2) for details.

[0063] h′ i =σ(Σa ij W j ); (2)

[0064] In the formula, h′ i represents the output vector of the i-th node; σ represents the activation function; a ij is the attention coefficient of the i-th node and its neighbor node j; W is the weight parameter matrix; h j is the output vector matrix of node j.

[0065] It should be noted that by introducing the attention mechanism, the problem that the graph convolutional neural network model can only assign the same weights to adjacent intersection points of the same order is solved, which makes the graph neural network model have better generalization ability to improve the accuracy of user sensitive data recognition.

[0066] As a preferred solution, the graph convolutional neural network model includes: an input layer, a graph convolution layer, a fully connected layer, a batch normalization layer, a spatiotemporal convolution layer, an average pooling layer, a maximum pooling layer and an output layer;

[0067] Among them, the spatiotemporal convolution layer is used to perform convolution operations on the spatiotemporal data of each node; the spatiotemporal data of each node is the spatiotemporal feature data formed by introducing the time dimension and space dimension into the node features of each node.

[0068] Please refer to Figure 2 , is a structural diagram of a virtual power plant user sensitive data identification system provided by an embodiment of the present invention, the system includes a data acquisition module M1 and a sensitive data identification module M2, and the specific details of each module are as follows:

[0069] The data acquisition module M1 is used to acquire the data graph of the users to be identified in the virtual power plant;

[0070] The sensitive data identification module M2 is used to input the user data graph to be identified into a pre-built sensitive data identification model, so that the sensitive data identification model can identify and process the user data graph to be identified and output the sensitivity level of the user data graph to be identified as the sensitive data identification result; wherein, the sensitive data identification model is constructed by introducing a graph attention mechanism on the basis of the graph convolution layer of the graph convolutional neural network model; the graph convolution layer that introduces the graph attention mechanism is used to assign weights to the nodes of the user data graph to be identified that is input into the sensitive data identification model based on the attention mechanism and combined with the attention coefficients of adjacent nodes, so as to extract the node features in the user data graph to be identified.

[0071] As a preferred solution, the data acquisition module M1 specifically includes a data acquisition unit 11 and a data graph construction unit 12, and the details of each unit are as follows:

[0072] The data collection unit 11 is used to collect virtual power plant user data of the virtual power plant in a preset time period in real time;

[0073] The data graph construction unit 12 is used to extract keywords from the virtual power plant user data, and use the extracted keywords as nodes, and the contextual relationship between each keyword as the edge between each node, and construct the user data graph to be identified of the virtual power plant based on all nodes and the edges between each node.

[0074] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the system described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0075] An embodiment of the present invention also provides a computer-readable storage medium, which includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a method for identifying sensitive data of virtual power plant users as described in Example 1.

[0076] An embodiment of the present invention also provides a terminal, including a processor, a memory, and a computer program stored in the memory; wherein the computer program can be executed by the processor to implement a virtual power plant user sensitive data identification as described in Embodiment 1.

[0077] Preferably, the computer program can be divided into one or more modules / units (such as computer program, computer program), one or more modules / units are stored in the memory and executed by the processor to complete the present invention. One or more modules / units can be a series of computer program instruction segments that can complete specific functions, and the instruction segments are used to describe the execution process of the computer program in the terminal.

[0078] The processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or the processor can also be any conventional processor. The processor is the control center of the terminal, and various parts of the terminal are connected using various interfaces and lines.

[0079] The memory mainly includes a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function, etc., and the data storage area can store related data, etc. In addition, the memory can be a high-speed random access memory, or a non-volatile memory, such as a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, and a flash card (Flash Card), etc., or the memory can also be other volatile solid-state storage devices.

[0080] It should be noted that the above-mentioned terminal may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that the above-mentioned terminal is merely an example and does not constitute a limitation on the terminal. It may include more or fewer components, or a combination of certain components, or different components.

[0081] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0082] The present invention provides a method, system, medium and terminal for identifying sensitive data of virtual power plant users. A graph attention mechanism is introduced on the basis of the graph convolution layer of a graph convolutional neural network model to form a sensitive data identification model, and a graph convolution layer of the graph attention mechanism is introduced to assign weights to each node of a user data graph to be identified that is input into the sensitive data identification model based on the attention mechanism and in combination with the attention coefficients of adjacent nodes, so as to extract node features in the user data graph to be identified. By the above means, a graph attention mechanism is introduced into the graph convolutional neural network model, which can change the defect that the graph convolutional neural network model can only assign the same weights to adjacent intersection points of the same order, so that the graph neural network model has better generalization ability and improves the accuracy of user sensitive data identification. After completing the optimization of the graph convolution layer of the graph convolution neural network model to obtain the sensitive data recognition model, the user data graph to be identified of the virtual power plant is input into the sensitive data recognition model, so that the sensitive data recognition model can identify and process the user data graph to be identified and output the sensitivity level of the user data graph to be identified as the sensitive data recognition result, so that the operation and maintenance management personnel can timely and accurately grasp the data security of the virtual power plant and provide data support for the preparation of optimization measures for the virtual power plant. In addition, through the graph convolution layer, the model can use the node and edge information in the graph structure to capture the contextual relationship between nodes, so that the model can more accurately identify and predict complex relationships and patterns, especially in dynamically changing scenarios.

[0083] The specific embodiments described above further illustrate the purpose, technical solutions and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for identifying sensitive data of virtual power plant users, characterized in that: include: Obtaining a user data graph to be identified in the virtual power plant; Inputting the user data graph to be identified into a pre-built sensitive data identification model, so that the sensitive data identification model performs identification processing on the user data graph to be identified and outputs the sensitivity level of the user data graph to be identified as a sensitive data identification result; Among them, the sensitive data identification model is constructed by introducing a graph attention mechanism on the basis of the graph convolution layer of the graph convolutional neural network model; the graph convolution layer introducing the graph attention mechanism is used to assign weights to the nodes of the user data graph to be identified that is input into the sensitive data identification model based on the attention mechanism and combined with the attention coefficients of adjacent nodes, so as to extract the node features in the user data graph to be identified.

2. A method for identifying sensitive data of virtual power plant users according to claim 1, characterized in that: The extraction of node features in the to-be-identified user data graph is specifically as follows: Using the graph attention mechanism, all nodes of the user data graph to be identified that is input into the sensitive data identification model are calculated to obtain the correlation between the nodes; Traversing each node of the to-be-identified user data graph, performing softmax normalization processing on the correlation between the current traversal node and the adjacent nodes of the current traversal node, and obtaining the attention coefficient corresponding to the current traversal node; According to the preset attention mechanism algorithm, based on the attention coefficient corresponding to the current traversal node and the output vector matrix of the adjacent nodes of the current traversal node, the output vector of the current traversal node is calculated as the node feature corresponding to the current traversal node.

3. A method for identifying sensitive data of virtual power plant users according to claim 2, characterized in that: According to the preset attention mechanism algorithm, the output vector of the current traversal node is calculated according to the attention coefficient corresponding to the current traversal node and the output vector matrix of the neighboring nodes of the current traversal node as the node feature corresponding to the current traversal node, specifically: h′ i =σ(∑a ij Wh j ); In the formula, h′ i represents the output vector of the i-th node; σ represents the activation function; a ij is the attention coefficient of the i-th node and its neighbor node j; W is the weight parameter matrix; h j is the output vector matrix of node j.

4. A method for identifying sensitive data of virtual power plant users according to claim 2, characterized in that: The expression of the graph attention mechanism is specifically: Attention(Q,S)=∑<Q,S i >·S i ; In the formula, S represents the user data graph to be identified in the virtual power plant, Q represents the sensitive data graph of known users in the virtual power plant, and S i Represents the i-th information in the user data graph to be identified of the virtual power plant.

5. A method for identifying sensitive data of virtual power plant users according to claim 1, characterized in that: The graph convolutional neural network model includes: an input layer, a graph convolution layer, a fully connected layer, a batch normalization layer, a spatiotemporal convolution layer, an average pooling layer, a maximum pooling layer and an output layer; Among them, the spatiotemporal convolution layer is used to perform convolution operations on the spatiotemporal data of each node; the spatiotemporal data of each node is spatiotemporal feature data formed by introducing time dimension and space dimension into the node features of each node.

6. A method for identifying sensitive data of virtual power plant users according to claim 1, characterized in that: The method of obtaining the to-be-identified user data map of the virtual power plant is specifically as follows: Collect virtual power plant user data in real time within a preset time period; Keywords are extracted from the user data of the virtual power plant, and the extracted keywords are used as nodes. The contextual relationships between the keywords are used as edges between the nodes. Based on all nodes and the edges between the nodes, a user data graph to be identified for the virtual power plant is constructed.

7. A virtual power plant user sensitive data identification system, characterized in that: include: A data acquisition module, used to acquire a data graph of users to be identified in a virtual power plant; A sensitive data identification module is used to input the user data graph to be identified into a pre-built sensitive data identification model, so that the sensitive data identification model identifies and processes the user data graph to be identified and outputs the sensitivity level of the user data graph to be identified as a sensitive data identification result; wherein, the sensitive data identification model is constructed by introducing a graph attention mechanism on the basis of a graph convolutional layer of a graph convolutional neural network model; the graph convolutional layer that introduces the graph attention mechanism is used to assign weights to the nodes of the user data graph to be identified that is input into the sensitive data identification model based on the attention mechanism and in combination with the attention coefficients of adjacent nodes, so as to extract the node features in the user data graph to be identified.

8. A virtual power plant user sensitive data identification system as claimed in claim 7, characterized in that: The data acquisition module specifically includes: A data collection unit, used to collect virtual power plant user data of the virtual power plant in a preset time period in real time; A data graph construction unit is used to extract keywords from the user data of the virtual power plant, and use the extracted keywords as nodes, and the contextual relationship between each keyword as the edge between each node, and construct the user data graph to be identified of the virtual power plant based on all nodes and the edges between each node.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program; wherein, when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a method for identifying sensitive data of virtual power plant users as described in any one of claims 1 to 6.

10. A terminal, characterized in that: It comprises a processor, a memory and a computer program stored in the memory; wherein the computer program can be executed by the processor to implement a method for identifying sensitive data of virtual power plant users as described in any one of claims 1 to 6.

Images