CN-121998409-A - Power equipment fault risk analysis method based on multidimensional feature fusion

CN121998409ACN 121998409 ACN121998409 ACN 121998409ACN-121998409-A

Abstract

The application provides a power equipment fault risk analysis method based on multidimensional feature fusion, which comprises the steps of collecting time sequence monitoring data of a plurality of nodes in a power system, intercepting T continuous time steps for each time of the time sequence monitoring data of each node by a sliding window to form a window section to be processed, merging the time sequence monitoring data and a time stamp for each time step in the window section to be processed to generate an enhanced feature vector to obtain a time sequence feature matrix formed by the T enhanced feature vectors, constructing a node feature matrix based on the time sequence feature matrix of each node, constructing a power information transmission link topological graph based on the relation among the plurality of nodes in the power system, further constructing a graph adjacent matrix, inputting the node feature matrix and the graph adjacent matrix into a preset fault risk analysis model to determine fault nodes, and outputting fault risk influence spreading range nodes by adopting an improved ant colony algorithm.

Inventors

DOU JIAN
LIU XUAN
QIE SHUANG
HAN YUE
HU CHEN

Assignees

中国电力科学研究院有限公司
国网江西省电力有限公司供电服务管理中心

Dates

Publication Date: 20260508
Application Date: 20251225

Claims (10)

1. The power equipment fault risk analysis method based on multidimensional feature fusion is characterized by comprising the following steps of: collecting time sequence monitoring data of a plurality of nodes in a power system, wherein the nodes represent different electric equipment in the power system; For time sequence monitoring data of each node, intercepting T continuous time steps by a sliding window each time to form a window segment to be processed, generating an enhanced feature vector by fusing the time sequence monitoring data and a time stamp for each time step in the window segment to be processed, and converting the window segment to be processed into a time sequence feature matrix formed by the T enhanced feature vectors; Extracting node feature vectors of each node based on the time sequence feature matrix of each node, combining the node feature vectors of all the nodes to construct a node feature matrix, constructing an electricity consumption transmission link topological graph based on physical connection and communication relations among the plurality of nodes in the power system, and constructing a graph adjacent matrix based on the topological graph; inputting the node characteristic matrix and the graph adjacent matrix into a preset fault risk analysis model to determine fault nodes; and based on the topological graph, simulating a propagation path of the fault risk by adopting an improved ant colony algorithm by taking the fault node as a starting point, and identifying and outputting nodes of which the fault risk affects the propagation range.
2. The method for analyzing risk of power equipment failure based on multidimensional feature fusion according to claim 1, wherein after the time sequence monitoring data of a plurality of nodes in the power system is collected, the method further comprises: the time sequence monitoring data of each node comprises a plurality of monitoring index values collected at a plurality of time points; extracting values at all time points of any one of the plurality of monitoring indexes to form an original index sequence, and carrying out standardization processing on the original index sequence by adopting a Z-score standardization method to obtain a standardized original index sequence; Obtaining standardized time sequence monitoring data of each node based on the standardized original index sequences of all monitoring indexes in each node; The corresponding code is used to determine the position of the object, And for the standardized time sequence monitoring data of each node, intercepting T continuous time steps by a sliding window each time to form a window segment to be processed, and then fusing the standardized time sequence monitoring data and a time stamp to generate an enhanced feature vector for each time step in the window segment to be processed, so as to convert the window segment to be processed into a time sequence feature matrix formed by the T enhanced feature vectors.
3. The method for analyzing the risk of power equipment failure based on multidimensional feature fusion according to claim 2, wherein the step of fusing standardized time series monitoring data and a time stamp to generate an enhanced feature vector comprises the steps of: And for each time step in the window section, splicing the standardized time sequence monitoring data corresponding to each time step, and the starting time stamp and the ending time stamp of data acquisition to form the enhancement feature vector corresponding to each time step.
4. The method for analyzing the risk of power equipment failure based on multidimensional feature fusion according to claim 1, wherein the extracting the node feature vector of each node based on the time sequence feature matrix of each node comprises: And carrying out mean pooling or maximum pooling on the enhanced feature vectors of all time steps in the time sequence feature matrix of each node, and taking the pooling result as the corresponding node feature vector.
5. The method for analyzing the risk of power equipment failure based on multidimensional feature fusion according to claim 1, wherein the constructing a graph adjacency matrix based on the topological graph comprises: Generating a basic adjacency matrix based on the connection relation between nodes in the topological graph; and adding an identity matrix on the basis of the basic adjacent matrix to obtain a graph adjacent matrix.
6. The power equipment fault risk analysis method based on multidimensional feature fusion according to claim 1, wherein the preset fault risk analysis model comprises a time sequence convolution network, a multi-layer graph neural network, an attention mechanism module, a hybrid expert network and a fault node determination module; the time sequence convolution network is used for extracting a long-short-period time sequence change mode of each node based on the node characteristic matrix and outputting a time sequence characteristic representation of each node; The multi-layer graph neural network is used for collecting the neighbor node characteristics of each node based on the node characteristic matrix and the graph adjacency matrix and outputting multi-layer spatial characteristic representation of each node; The attention mechanism module is used for dynamically fusing the time sequence characteristic representation and the multi-layer space characteristic representation of each node by using an attention mechanism to generate a fused characteristic representation of each node; the multi-layer hybrid expert network is used for calculating multi-layer anomaly probability of each node based on the fusion characteristic representation of each node. The fault node determining module is used for outputting the abnormal probability value of each node based on the multi-layer abnormal probability of each node by combining dynamic route weighted summation, and taking the node with the highest abnormal probability value as the fault node.
7. The method for analyzing the risk of failure of the power equipment based on the multi-dimensional feature fusion according to claim 1, wherein the steps of using the failure node as a starting point, simulating a propagation path of the failure risk by adopting an improved ant colony algorithm based on the topological graph, and identifying and outputting the node of the failure risk influence spreading range include the steps of: S51, initially releasing the artificial ants at the fault nodes, endowing each side in the topological graph with the same initial pheromone concentration value, and limiting the artificial ants to move along the preset data transmission path direction on the topological graph; S52, in each iteration, the artificial ants move from the current node to an inaccessible and reachable neighbor node according to the path selection probability, wherein the path selection probability is calculated by the pheromone concentration value on the corresponding side and heuristic information, the heuristic information is determined based on the abnormal probability calculation of the current node and the neighbor node of the current node, and the heuristic information is used for guiding the artificial ants to move to the region with high abnormal probability; S53, after the path selection of the artificial ants is completed in each round of iteration, globally updating the pheromone concentration of each side in the topological graph, wherein the updating process comprises the steps of volatilizing the pheromone concentration of all sides according to a preset volatilizing coefficient in proportion, simulating the natural attenuation of information; S54, repeatedly executing the steps S52 to S53 until the preset iteration times are reached, determining the node connected with the edge with the final pheromone concentration exceeding the preset threshold value as the node with the fault risk influence spreading range, and outputting the node.
8. A power equipment fault risk analysis system based on multidimensional feature fusion, the system comprising: the data acquisition and processing module is used for acquiring time sequence monitoring data of a plurality of nodes in the power system, wherein the plurality of nodes represent different electric equipment in the power system; The time sequence feature construction module is used for intercepting T continuous time steps each time for the time sequence monitoring data of each node by the sliding window to form a window segment to be processed, generating an enhanced feature vector by fusing the time sequence monitoring data and the time stamp for each time step in the window segment to be processed, and converting the window segment to be processed into a time sequence feature matrix formed by the T enhanced feature vectors; The node characteristic and graph structure modeling module is used for extracting node characteristic vectors of each node based on the time sequence characteristic matrix of each node, combining the node characteristic vectors of all the nodes to construct a node characteristic matrix, constructing an electricity information transmission link topological graph based on physical connection and communication relations among the plurality of nodes in the electric power system, and constructing a graph adjacent matrix based on the topological graph; The fault risk analysis model module is used for inputting the node characteristic matrix and the graph adjacency matrix into a preset fault risk analysis model to determine fault nodes; And the risk spread analysis module is used for simulating a propagation path of the fault risk by adopting an improved ant colony algorithm based on the topological graph by taking the fault node as a starting point, and identifying and outputting nodes of which the fault risk affects a spread range.
9. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor performs the steps of the power device fault risk analysis method based on multi-dimensional feature fusion as claimed in any one of claims 1 to 7.
10. A computer readable storage medium, characterized in that the computer readable storage medium comprises a storage computer program or instructions which, when executed, cause the method of any of claims 1-7 to be performed.

Description

Power equipment fault risk analysis method based on multidimensional feature fusion Technical Field The application relates to the technical field of smart grids, in particular to a power equipment fault risk analysis method, a system, electronic equipment and a storage medium based on multidimensional feature fusion. Background With the rapid advancement of new power system construction, remote data acquisition devices deployed in power networks, such as smart meters, terminals, concentrators, etc., exhibit explosive growth. These devices are widely distributed in areas, cover complex physical environments from cities to villages, and are exposed to diverse climates and electromagnetic interference for a long period of time, so that their operational status monitoring and operation maintenance face unprecedented challenges. The traditional operation and maintenance mode relying on manual inspection and post-inspection is difficult to adapt to new situations with huge equipment quantity and high requirements on fault recession and instantaneity, so that the operation and maintenance efficiency is low, and the labor and time cost is also increased sharply. Meanwhile, the system gathers operation data from mass equipment every day, including multidimensional information such as voltage, current, communication state, event log and the like, and has huge data volume and relatively low value density. Because the data has the characteristics of multiple sources, isomerism, strong time sequence and the like, the hidden equipment health state degradation trend and early fault symptoms in the data are difficult to deeply mine by simply relying on threshold value alarming or traditional statistical analysis means. The insufficient analysis capability leads to perception lag of potential risks of equipment, and accurate positioning and early warning of faults cannot be realized, so that fault discovery often depends on caused business influences, such as measurement data acquisition failure, abnormal electric quantity and the like, and acquisition success rate and data quality are directly influenced. Disclosure of Invention The embodiment of the application provides a power equipment fault risk analysis method based on multidimensional feature fusion, which realizes the identification of the spreading range of the influence of the power equipment fault risk, so as to support the real-time monitoring, abnormal identification and efficient tracing of the remote power equipment fault and promote the intelligent operation and maintenance of a novel power system. In order to achieve the above purpose, the application adopts the following technical scheme: in a first aspect, the present application provides a method for analyzing risk of failure of an electrical device based on multidimensional feature fusion, the method comprising: collecting time sequence monitoring data of a plurality of nodes in the power system, wherein the plurality of nodes represent different electric equipment in the power system; For the time sequence monitoring data of each node, the sliding window intercepts T continuous time steps each time to form a window segment to be processed, and then each time step in the window segment to be processed is fused with the time sequence monitoring data and a time stamp to generate an enhanced feature vector, and the window segment to be processed is converted into a time sequence feature matrix formed by the T enhanced feature vectors; Extracting node feature vectors of each node based on the time sequence feature matrix of each node, combining the node feature vectors of all nodes to construct a node feature matrix, constructing an electricity consumption transmission link topological graph based on physical connection and communication relations among a plurality of nodes in an electric power system, and constructing a graph adjacency matrix based on the topological graph; inputting the node characteristic matrix and the graph adjacent matrix into a preset fault risk analysis model, and determining fault nodes; Based on a topological graph, an improved ant colony algorithm is adopted to simulate a propagation path of fault risks by taking a fault node as a starting point, and nodes in a fault risk influence spreading range are identified and output. In a second aspect, the present application provides a power equipment fault risk analysis system based on multidimensional feature fusion, the system comprising: the data acquisition and processing module is used for acquiring time sequence monitoring data of a plurality of nodes in the power system, wherein the plurality of nodes represent different electric equipment in the power system; The time sequence feature construction module is used for intercepting T continuous time steps each time for the time sequence monitoring data of each node by the sliding window to form a window segment to be processed, generating an enhanced feature vector by fusing the t