A power protection method and device based on intelligent network

Abstract

The present invention discloses an intelligent network-based power protection method and device, which relates to the field of power protection technology, and comprises the following steps: obtaining and preprocessing the operating status data of power equipment in a power grid and the characteristic data of the power lines connecting the power equipment in the power grid in the district; constructing a power flow diagram based on the topological structure of the power grid, and dividing the power grid into G subnets based on the power flow diagram; taking the G subnets as graph nodes and the power lines connecting the power equipment as edges to obtain G power grid subnet graphs, and using a preset convolutional neural network to obtain G power grid subnet state feature vectors; inputting the G power grid subnet state feature vectors and the operating status data of the power equipment in the power grid and the characteristic data of the power lines connecting the power equipment in the district power grid into a preset GCNN-based power grid state prediction model to obtain a power grid state prediction value; and performing power protection early warning based on the power grid state prediction value.

Description

Technical Field

[0001] The present invention relates to the technical field of power protection, and more specifically, to a power protection method and device based on an intelligent network. Background Art

[0002] Power protection is an important part of ensuring the safe, stable and reliable operation of the power system. Its core purpose is to detect and isolate faults in time when power equipment fails or is abnormal, so as to protect the overall function of the system and avoid equipment damage and casualties. With the continuous growth of global power demand, the power system is facing unprecedented pressure. On the one hand, this is due to the accelerated urbanization process, which has led to a continuous increase in demand for electricity; on the other hand, with the development of the economy, industrial and commercial electricity consumption has increased significantly, further increasing the burden on power supply.

[0003] Traditional power protection is mainly based on simple threshold comparison of current and voltage, which is easily affected by interference and false faults (such as current fluctuations, load changes, etc.), resulting in false operation or refusal to operate, thus affecting the safety and reliability of the power system. At the same time, traditional power protection methods are difficult to cope with complex power network environments. The phenomenon of information islands is prevalent, resulting in problems such as low accuracy of power warnings. In view of the above problems, the present invention proposes a solution. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an intelligent network-based power protection method and device, which solves the problem of low accuracy of traditional methods in power prediction and protection.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A power protection method based on an intelligent network comprises the following steps: obtaining and preprocessing the operating status data of power equipment in a power grid and the characteristic data of the power lines connecting the power equipment in the power grid; constructing a power flow diagram based on the topological structure of the power grid, and dividing the power grid into G subnets based on the power flow diagram; taking the G subnets as graph nodes and the power lines connecting the power equipment as edges to obtain G power grid subnet graphs, and inputting the node power data of the G power grid subnet graphs into a preset convolutional neural network for feature extraction to obtain G power grid subnet state feature vectors; inputting the G power grid subnet state feature vectors and the operating status data of the power equipment in the power grid and the characteristic data of the power lines connecting the power equipment in the power grid into a preset GCNN-based power grid state prediction model to obtain a power grid state prediction value; and performing a power protection warning based on the power grid state prediction value and a preset power grid state threshold.

[0007] In a preferred embodiment, the power flow graph is constructed based on the topological structure of the power grid, and the power grid is divided into G subnets based on the power flow graph, specifically: based on the topological structure of the power grid, each power device in the power grid is used as a graph node, and the power lines connecting the power devices are used as edges to construct a power grid graph structure, and the power grid adjacency matrix of the power grid is calculated, and the elements in the power grid adjacency matrix of the power grid are 1 when two power devices are adjacent, and 0 when they are not adjacent; the power flow between the power devices is calculated based on the power flow variance, and the power flow between the power devices is used as the weight of the edge in the power grid graph structure to obtain the power grid edge weight matrix; based on the power grid adjacency matrix and the power grid edge weight matrix of the power grid, a power flow graph is constructed; based on the power flow graph, the Louvain algorithm is used to divide the power grid into G subnets.

[0008] In a preferred embodiment, based on the power flow graph, the Louvain algorithm is used to divide the power grid into G subnets, specifically: each node in the power flow graph is regarded as an original independent power grid community; the power grid modularity is defined as an evaluation index for the division of the original independent power grid community; the power grid modularity change value of the original power grid community of each node in the power flow graph moving to the corresponding adjacent node is calculated; based on the power grid modularity change value, each power grid community is regarded as a new power grid super node, and a new power grid graph structure is constructed, and the edge weight of the new power grid graph structure is the sum of the power flow rates between the original independent power grid communities in the power flow graph; based on the new power grid graph structure, a new power grid adjacency matrix and a power grid weight matrix are constructed to obtain G subnets.

[0009] In a preferred embodiment, a grid modularity change value of an original grid community of each node in the power flow diagram moved to a corresponding adjacent node is calculated, specifically: the grid modularity of the original grid community of each node in the power flow diagram is calculated; the grid modularity of the original grid community of each node in the power flow diagram is calculated to move to a corresponding adjacent node to obtain the grid modularity after movement; the grid modularity after movement is subtracted from the grid modularity of the original grid community to obtain a grid modularity change value; when the grid modularity change value is greater than 0, each node in the power flow diagram is moved to the original grid community corresponding to the adjacent node.

[0010] In a preferred embodiment, the G power grid subnet state feature vectors and the operating status data of the power grid power equipment and the feature data of the power lines connecting the power grid power equipment in the district are input into a preset GCNN-based power grid state prediction model to obtain a power grid state prediction value, specifically: the GCNN-based power grid state prediction model includes an input layer, k graph convolution layers with a graph attention mechanism, an LSTM layer and a fully connected layer; the G power grid subnet state feature vectors are weighted fused to obtain a power grid state fusion feature vector; the power grid state fusion feature vector and the operating status data of the power grid power equipment and the feature data of the power lines connecting the power grid power equipment in the district are input into the GCNN-based power grid state prediction model for prediction to obtain a power grid state prediction value.

[0011] In a preferred embodiment, the k graph convolution layers that add the graph attention mechanism are specifically: the graph attention mechanism assigns preset weights to the adjacent nodes of each subnet node in the power grid graph structure to obtain a weighted power grid state feature vector; and the weighted power grid state feature vector is used as the input of the LSTM layer.

[0012] The device of any one of the power protection methods based on intelligent networks is characterized by comprising a data acquisition module, a subnet division module, a subnet state feature extraction module, a power grid state prediction module, and a power protection module, and there are connections between the modules:

[0013] A data acquisition module, which acquires the operating status data of the power equipment of the power grid, the characteristic data of the power equipment and the characteristic data of the power lines connecting the power equipment of the regional power grid, and performs preprocessing;

[0014] The subnet division module constructs a power flow diagram based on the topological structure of the power grid, and divides the power grid into G subnets based on the power flow diagram;

[0015] A subnet state feature extraction module, which uses the G subnets as graph nodes and the power lines connecting the power equipment as edges to obtain G power grid subnet graphs, and inputs the node power data of the G power grid subnet graphs into a preset convolutional neural network for feature extraction to obtain G power grid subnet state feature vectors;

[0016] A power grid state prediction module, which inputs the G power grid subnet state feature vectors and the operating state data of the power grid power equipment into a preset GCNN-based power grid state prediction model to obtain a power grid state prediction value;

[0017] The power protection module takes corresponding power protection measures based on the power grid state prediction value and the preset threshold value.

[0018] The technical effects and advantages of the power protection method and device based on intelligent network of the present invention are as follows:

[0019] 1. The present invention obtains and pre-processes the operating status data of the power equipment of the power grid and the characteristic data of the power lines, builds a power flow diagram in combination with the topological structure of the power grid, and divides the power grid into GGG subnets. This process helps to perceive the real-time status of the power grid more accurately. Dividing the power grid into multiple subnets can simplify the analysis of complex systems, allowing each subnet to independently characterize the local operating status. By inputting the power data of the subnet nodes into the convolutional neural network (CNN), key features can be effectively extracted, so that the model can capture the power flow relationship between devices, equipment load conditions, etc. Subsequently, the state of the entire power grid is predicted using a model based on a graph convolutional neural network (GCNN), further enhancing the global perception capability of the power grid. GCNN can improve the accuracy of prediction by aggregating the information of each subnet and taking into account the characteristics of device connections and topological structures, especially when dealing with the complex nonlinear characteristics of the power system, thereby helping to detect power grid anomalies earlier and issue early warnings in a timely manner. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 This is a schematic diagram of the structure of an intelligent network-based power protection method of the present invention.

[0021] Figure 2 The present invention is a schematic diagram of the structure of an intelligent network-based power protection method. DETAILED DESCRIPTION

[0022] 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.

[0023] Embodiment 1, Figure 1 The present invention provides a power protection method based on an intelligent network, comprising the following steps:

[0024] S1, obtaining the operating status data of the power equipment of the power grid and the characteristic data of the power lines connecting the power equipment of the power grid in the area and performing preprocessing;

[0025] In this example, the operating status data of the power grid power equipment and the characteristic data of the power line connecting the power grid power equipment are specifically:

[0026] The operating status data of the power grid equipment includes current and voltage data, power data, power grid frequency data, and equipment failure data; the characteristic data of the power lines connecting the power grid equipment includes line impedance, line current and voltage, transmission power, and harmonic content data.

[0027] It should be noted that obtaining the operating status data of power equipment in the power grid and the characteristic data of the power lines connecting the power equipment in the power grid is crucial for the safe, stable and efficient operation of the power grid. First, the operating status data (current, voltage, power, temperature, etc.) of power equipment (such as transformers, generators, circuit breakers, etc.) can reflect the load and health of the equipment in real time, which helps to timely discover problems such as overload, short circuit or equipment aging, so as to take preventive maintenance or protection measures to avoid equipment failure or power outage accidents. Secondly, the characteristic data of power lines (such as impedance, length, temperature, load rate, harmonics, etc.) can reveal the transmission capacity and line status of power in the network, help identify power loss, overload operation or line abnormality, and ensure that electric energy can be transmitted reliably and safely. Through real-time monitoring and analysis of these data, the power system can predict faults, optimize scheduling and improve power quality, ensure the efficient operation, stability and reliability of the power grid, and provide decision-making basis for power grid operators to improve the intelligent management level of the power system.

[0028] S2, based on the topological structure of the power grid, construct a power flow diagram, and divide the power grid into G subgrids based on the power flow diagram;

[0029] In this example, the power flow diagram is constructed based on the topological structure of the power grid, and the power grid is divided into G subnets based on the power flow diagram, specifically:

[0030] Based on the topological structure of the power grid, each power device in the power grid is taken as a graph node, and the power lines connecting the power devices are taken as edges to construct a power grid graph structure, and a power grid adjacency matrix of the power grid is calculated, wherein the elements in the power grid adjacency matrix of the power grid have a value of 1 when two power devices are adjacent, and a value of 0 when they are not adjacent;

[0031] Calculating the power flow between the power devices based on the power flow variance, and using the power flow between the power devices as the weight of the edge in the power grid graph structure to obtain a power grid edge weight matrix;

[0032] constructing a power flow graph based on a grid adjacency matrix and a grid edge weight matrix of the grid;

[0033] Based on the power flow diagram, the power grid is divided into G subgrids using the Louvain algorithm.

[0034] In this example, the topological structure based on the power grid is constructed by taking each power device in the power grid as a graph node and the power lines connecting the power devices as edges, constructing a power grid graph structure, and calculating the power grid adjacency matrix of the power grid. The elements in the power grid adjacency matrix of the power grid are 1 when two power devices are adjacent and 0 when they are not adjacent, specifically:

[0035] Based on the topological structure of the power grid, each power device in the power grid is taken as a graph node, the power lines connecting the power devices are taken as edges, the power grid graph structure is constructed, and the adjacency matrix of the power grid is calculated, which can bring multiple benefits. First, the power grid graph structure clearly shows the connection relationship between power devices. The adjacency matrix can intuitively represent the topological structure of the power system, reflecting which devices are interrelated, which helps to understand the path of power flow and the overall layout of the network. Secondly, the adjacency matrix can be used as the basic input for power system data analysis and modeling, providing structured data for advanced algorithms such as graph neural networks (GNNs), and then used for tasks such as power grid state prediction, fault detection, and load distribution. In addition, this graph representation method facilitates the segmentation and modular management of the network, helps identify key nodes or vulnerable links in the power grid, and improves the operational reliability and risk resistance of the power grid. By constructing the power grid graph structure and adjacency matrix, the power system can more efficiently perform topology optimization, fault location, and intelligent scheduling, laying the foundation for intelligent management and automated protection of the power grid.

[0036] In this example, the power flow between the power devices is calculated based on the power flow variance, specifically:

[0037]

[0038] in, and are the injected active power and reactive power of the ith power equipment, respectively. and is the voltage amplitude of power equipment i and power equipment j, and are the real and imaginary parts of the admittance between power equipment i and power equipment j, respectively. and are the voltage phase angles of power equipment i and power equipment j respectively, sin() is the sine function, and cos() is the cosine function.

[0039] It should be noted that power flow analysis is an important tool for power grid operation. By calculating the power flow between power equipment, the actual flow direction and distribution of power in the power grid can be determined. By calculating the power flow variance in different time periods, the volatility and stability of power flow can be identified, and the load fluctuations and power exchange rules between power equipment can be revealed. This method is particularly suitable for monitoring power bottlenecks, overload nodes or potential fault points in the power grid, because abnormal fluctuations in power flow usually indicate equipment overload or unbalanced power lines. Through power flow variance calculation, the operation and scheduling of power equipment can also be optimized to ensure that power flows more smoothly and efficiently in the system, reducing energy loss and power quality problems. This calculation method provides a quantitative basis for monitoring the state of the power grid, predicting the operating risks of the power system, and optimizing the operation of the power grid. It is an important tool for power protection and fault diagnosis.

[0040] In this example, based on the power flow diagram, the Louvain algorithm is used to divide the power grid into G subgrids, specifically:

[0041] Each node in the power flow graph is regarded as an original independent power grid community;

[0042] defining grid modularity as an evaluation index for dividing the original independent grid communities;

[0043] Calculate the grid modularity change value of the original grid community where each node in the power flow graph moves to a corresponding adjacent node;

[0044] Based on the grid modularity change value, each grid community is used as a new grid super node, and a new grid graph structure is constructed, wherein the edge weight of the new grid graph structure is the sum of the power flow rates between the original independent grid communities in the power flow graph;

[0045] Based on the new power grid graph structure, a new power grid adjacency matrix and a power grid weight matrix are constructed to obtain G subnets.

[0046] It should be noted that using the Louvain algorithm to divide the power grid into G subnets is an effective power grid partitioning strategy that can significantly improve the management and operation efficiency of the power system. The Louvain algorithm is a community partitioning algorithm based on modularity optimization, which can divide the network into multiple modules according to the connection relationship and weight between nodes. In the power grid, by using power equipment as nodes and power lines as edges, the Louvain algorithm can divide the power grid into several subnets according to the strength of power flow. In this way, devices with close power transmission relationships can be gathered in the same subnet to achieve regional management. This division helps to simplify the analysis of complex power grids and improve the controllability and maintenance efficiency of the power grid. For example, when a fault occurs in an area, localized division can limit the scope of the fault impact, quickly isolate and handle the problem. In addition, the division of subnets can also optimize power dispatching and load distribution, and improve the operation efficiency and reliability of the power grid. The efficiency and flexibility of the Louvain algorithm make it particularly suitable for modular management of power grids.

[0047] In this example, the grid modularity change value of the original grid community of each node in the power flow diagram moving to the corresponding adjacent node is calculated, specifically:

[0048] Calculating the grid modularity of the original grid community of each node in the power flow graph;

[0049] Calculate the grid modularity of the original grid community of each node in the power flow graph moved to the corresponding adjacent node to obtain the grid modularity after the move;

[0050] Subtracting the modularity of the power grid after the movement from the modularity of the original power grid community to obtain a change value of the power grid modularity;

[0051] When the grid modularity change value is greater than 0, each node in the power flow graph is moved to the original grid community corresponding to the adjacent node.

[0052] In this example, when the grid modularity change value is greater than 0, each node in the power flow graph is moved to the original grid community corresponding to the adjacent node, specifically:

[0053] In the process of executing the Louvain algorithm, modularity is an important indicator to measure the quality of community division, which represents the difference between the connection density between nodes in the community and the connection density outside the community. When a node moves to its adjacent community, the new modularity change value ΔQ is calculated. If ΔQ is greater than 0, it means that the modularity of the community division increases after the node moves, that is, the quality of the community division is improved.

[0054] In this example, the definition of grid modularity as an evaluation index for the original independent grid community division is specifically:

[0055]

[0056] in, is the grid modularity, is the total weight of the edge, is the edge weight between node i and node j in the power flow graph, and are the degrees of node i and node j in the power flow graph, is the indicator function.

[0057] S3, taking the G subnets as graph nodes and the power lines connecting the power equipment as edges to obtain G power grid subnet graphs, and inputting the node power data of the G power grid subnet graphs into a preset convolutional neural network for feature extraction to obtain G power grid subnet state feature vectors;

[0058] In this example, the node power data of the G power grid subnet graphs are input into a preset convolutional neural network for feature extraction to obtain G power grid subnet state feature vectors, specifically:

[0059] Inputting the node power data of the GGG power grid subnetwork graphs into the preset convolutional neural network (CNN) for feature extraction can efficiently extract key state features from the complex data of the power grid. First, the node power data of each subnetwork graph, including voltage, current, power, etc., can be used as input features, and local features are extracted through the convolution layer of CNN. The convolution operation of CNN can effectively identify the association pattern between nodes and their adjacent nodes, extract the local state features of different nodes in the power system, and help capture the interaction between power equipment and the distribution law of power flow.

[0060] Secondly, the multi-layer structure of CNN allows for the extraction of higher-level and more abstract features layer by layer. As the data passes through multiple convolutional and pooling layers, the model is able to gradually extract high-level features that reflect the overall subnet health status and power balance from low-level features. Ultimately, the node features of each subnet are summarized into a state feature vector, representing the overall operating status of the entire subnet. These subnet feature vectors can be further used for global grid state prediction, fault detection, and dispatch optimization, improving the intelligent management capabilities of the power system.

[0061] S4, inputting the G grid subnet state feature vectors and the operating state data of the grid power equipment and the feature data of the power lines connecting the district grid power equipment into a preset GCNN-based grid state prediction model to obtain a grid state prediction value;

[0062] In this example, the G grid subnet state feature vectors and the operating state data of the grid power equipment and the feature data of the power lines connecting the regional grid power equipment are input into the preset GCNN-based grid state prediction model to obtain the grid state prediction value, specifically:

[0063] The GCNN-based power grid state prediction model includes an input layer, k graph convolution layers with graph attention mechanism, an LSTM layer and a fully connected layer;

[0064] Performing weighted fusion on the G power grid subnet state feature vectors to obtain a power grid state fusion feature vector;

[0065] The grid state fusion feature vector, the operating state data of the grid power equipment, and the feature data of the power lines connecting the district grid power equipment are input into the GCNN-based grid state prediction model for prediction to obtain a grid state prediction value.

[0066] In this example, the specific formula of the graph attention mechanism is:

[0067]

[0068] in, is the updated feature of node i, is the set of adjacent nodes of node i, is the attention weight between node i and node j, is the learnable weight matrix of the Lth layer, is the feature vector of node j at the Lth layer.

[0069] It should be noted that building k graph convolutional layers with graph attention mechanisms has many advantages, especially in processing graph-structured data such as complex power grids. First, the graph attention mechanism allows the model to dynamically assign different neighbor node weights to each node, thereby capturing the importance of different neighbors to the node. This is particularly critical in complex networks such as power grids, because the interaction between certain power devices has a much greater impact on the system state than other devices. Through the attention mechanism, the model can more accurately focus on neighbor nodes that have a greater impact on the node, thereby improving the efficiency of information utilization.

[0070] Secondly, adding multiple graph convolution layers can aggregate information from more distant neighbors layer by layer, which means that through the design of k layers, the model can consider local and global power network information at the same time, improving the perception and prediction capabilities of the entire network status. Multi-layer design can also enhance the expressiveness of the model, enabling it to capture deeper topological structure information. Combined with the attention mechanism, multi-layer graph convolution can effectively solve the problem of grid heterogeneity, improve the modeling effect of complex tasks such as power flow and load distribution, and ultimately improve the prediction accuracy and protection capabilities of the power system.

[0071] S5, performing power protection warning based on the grid state prediction value and the preset grid state threshold.

[0072] In this example, based on the grid state prediction value and the preset grid state threshold, a power protection early warning is performed, specifically:

[0073] When the grid status prediction value is close to the preset grid status threshold, the grid has a slight abnormality and may need attention;

[0074] When the grid status prediction value exceeds the preset grid status threshold, the grid is at great risk and partial power cuts or equipment switching may be required;

[0075] When the predicted value of the grid state seriously exceeds the preset grid state threshold, the power system may face failure and protective measures must be taken immediately, such as shedding loads and starting backup power supplies.

[0076] like Figure 2 The figure shows a functional module diagram of a power protection method based on an intelligent network. According to the functions to be implemented, the power protection method based on an intelligent network may include a data acquisition module, a subnet division module, a subnet state feature extraction module, a power grid state prediction module, and a power protection module. The module described in the invention may also be referred to as a unit, which refers to a series of computer program segments that can be executed by a processor of a personnel performance data processing device and can complete fixed functions, and are stored in a memory of the personnel performance data processing device.

[0077] In this embodiment, the functions of each module are as follows:

[0078] A data acquisition module, which acquires the operating status data of the power equipment of the power grid, the characteristic data of the power equipment and the characteristic data of the power lines connecting the power equipment of the regional power grid, and performs preprocessing;

[0079] The subnet division module constructs a power flow diagram based on the topological structure of the power grid, and divides the power grid into G subnets based on the power flow diagram;

[0080] A subnet state feature extraction module, which uses the G subnets as graph nodes and the power lines connecting the power equipment as edges to obtain G power grid subnet graphs, and inputs the node power data of the G power grid subnet graphs into a preset convolutional neural network for feature extraction to obtain G power grid subnet state feature vectors;

[0081] A power grid state prediction module, which inputs the G power grid subnet state feature vectors and the operating state data of the power grid power equipment into a preset GCNN-based power grid state prediction model to obtain a power grid state prediction value;

[0082] The power protection module takes corresponding power protection measures based on the power grid state prediction value and the preset threshold value.

[0083] In detail, the modules described in the intelligent network-based power protection method device described in the embodiment of the present invention adopt the same technical means as the intelligent network-based power protection method described in the accompanying drawings when used, and can produce the same technical effects, which will not be repeated here.

[0084] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0085] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0086] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0087] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0088] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0089] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A power protection method based on intelligent network, characterized in that: The following steps are involved: Acquire the operating status data of the power equipment of the power grid and the characteristic data of the power lines connecting the power equipment of the power grid and perform preprocessing; Based on the topological structure of the power grid, a power flow diagram is constructed, and the power grid is divided into G subgrids based on the power flow diagram; Based on the topological structure of the power grid, each power device in the power grid is taken as a graph node, and the power lines connecting the power devices are taken as edges to construct a power grid graph structure, and a power grid adjacency matrix of the power grid is calculated, wherein the elements in the power grid adjacency matrix of the power grid have a value of 1 when two power devices are adjacent, and a value of 0 when two power devices are not adjacent; Calculating the power flow between the power devices based on the power flow variance, and using the power flow between the power devices as the weight of the edge in the power grid graph structure to obtain a power grid edge weight matrix; constructing a power flow graph based on a grid adjacency matrix and a grid edge weight matrix of the grid; Based on the power flow diagram, the Louvain algorithm is used to divide the power grid into G subgrids, specifically: Each node in the power flow graph is regarded as an original independent power grid community; defining grid modularity as an evaluation index for dividing the original independent grid communities; The grid modularity change value of the original grid community where each node in the power flow graph moves to the corresponding adjacent node is calculated, specifically: Calculating the grid modularity of the original grid community of each node in the power flow graph; Calculate the grid modularity of the original grid community of each node in the power flow graph moved to the corresponding adjacent node to obtain the grid modularity after the move; Subtracting the modularity of the power grid after the movement from the modularity of the original power grid community to obtain a change value of the power grid modularity; When the grid modularity change value is greater than 0, each node in the power flow graph is moved to the original grid community corresponding to the adjacent node; Based on the grid modularity change value, each grid community is used as a new grid super node, and a new grid graph structure is constructed, wherein the edge weight of the new grid graph structure is the sum of the power flow rates between the original independent grid communities in the power flow graph; Based on the new power grid graph structure, construct a new power grid adjacency matrix and a power grid weight matrix to obtain G subnets; The G subnets are used as graph nodes, and the power lines connecting the power equipment are used as edges to obtain G power grid subnet graphs, and the node power data of the G power grid subnet graphs are input into a preset convolutional neural network for feature extraction to obtain G power grid subnet state feature vectors; Input the G grid subnet state feature vectors, the operating state data of the grid power equipment, and the feature data of the power lines connecting the grid power equipment into a preset GCNN-based grid state prediction model to obtain a grid state prediction value; Based on the grid state prediction value and the preset grid state threshold, a power protection early warning is performed.

2. The power protection method based on intelligent network according to claim 1, characterized in that: The G grid subnet state feature vectors, the operating state data of the grid power equipment, and the feature data of the power lines connecting the grid power equipment are input into a preset GCNN-based grid state prediction model to obtain a grid state prediction value, specifically: The GCNN-based power grid state prediction model includes an input layer, k graph convolution layers with graph attention mechanism, an LSTM layer and a fully connected layer; Performing weighted fusion on the G power grid subnet state feature vectors to obtain a power grid state fusion feature vector; The grid state fusion feature vector, the operating state data of the grid power equipment, and the feature data of the power lines connecting the grid power equipment are input into the GCNN-based grid state prediction model for prediction to obtain a grid state prediction value.

3. The intelligent network-based power protection method according to claim 2, characterized in that: The k graph convolutional layers that add the graph attention mechanism are specifically: The graph attention mechanism assigns preset weights to the neighboring nodes of each subnet node in the power grid graph structure to obtain a weighted power grid state feature vector; The weighted power grid state feature vector is used as the input of the LSTM layer.

4. The intelligent network-based power protection method according to claim 3, characterized in that: The power flow between the power equipment is calculated based on the power flow variance, specifically: in, and are the injected active power and reactive power of the ith power equipment, respectively. and is the voltage amplitude of power equipment i and power equipment j, and are the real and imaginary parts of the admittance between power equipment i and power equipment j, respectively. and are the voltage phase angles of power equipment i and power equipment j respectively, sin() is the sine function, and cos() is the cosine function.

5. The intelligent network-based power protection method according to claim 4, characterized in that: The definition of grid modularity as an evaluation index for the original independent grid community division is specifically: in, is the grid modularity, is the total weight of the edge, is the edge weight between node i and node j in the power flow graph, and are the degrees of node i and node j in the power flow graph, is the indicator function.

6. The intelligent network-based power protection method according to claim 5, characterized in that: The specific formula of the graph attention mechanism is: in, is the updated feature of node i, is the set of adjacent nodes of node i, is the attention weight between node i and node j, is the learnable weight matrix of the Lth layer, is the feature vector of node j at the Lth layer.

7. A device using a power protection method based on an intelligent network as claimed in any one of claims 1 to 6, characterized in that: It includes data acquisition module, subnet division module, subnet status feature extraction module, power grid status prediction module, power protection module, and there are connections between modules: A data acquisition module, used to acquire and pre-process the operating status data of the power equipment of the power grid and the characteristic data of the power lines connecting the power equipment of the power grid; A subnet division module is used to construct a power flow diagram based on the topological structure of the power grid, and divide the power grid into G subnets based on the power flow diagram; A subnet state feature extraction module, used to use the G subnets as graph nodes and the power lines connecting the power equipment as edges to obtain G power grid subnet graphs, and input the node power data of the G power grid subnet graphs into a preset convolutional neural network for feature extraction to obtain G power grid subnet state feature vectors; A power grid state prediction module, used to input the G power grid subnet state feature vectors and the operating state data of the power grid power equipment and the feature data of the power lines connecting the power grid power equipment into a preset GCNN-based power grid state prediction model to obtain a power grid state prediction value; The power protection module is used to perform power protection early warning based on the power grid state prediction value and a preset power grid state threshold.

Images