Research On Anomaly Monitoring Algorithm Of Smart Grid Based On Graph Neural Network

In recent years,with the continuous improvement of the intelligent level of the power grid and the continuous expansion of the scale,the existing smart grid anomaly detection methods and methods are limited by the complex power grid environment,which puts forward new requirements for the anomaly monitoring of the smart grid and faces great challenges.Owing to the development of graph signal processing theory,the problem of signal processing in irregular field has been paid more attention and studied.Generally speaking,the research and processing of structured data with certain arrangement rules(such as time series and images)has a complete theoretical system and processing algorithm in the traditional signal processing theory.As unstructured power grid data,the classical signal analysis method has been difficult to apply to complex power grid models.Therefore,this paper proposes a smart grid anomaly detection algorithm based on graph neural network.Firstly,the smart grid is constructed as a graph model,and on this basis,the graph neural network detection model is further constructed,so as to realize an effective anomaly detection algorithm.This paper studies the anomaly monitoring of smart grid in DC mode.The anomaly monitoring algorithm based on graph filter and the anomaly monitoring algorithm based on graph convolution neural network are designed respectively.Simulation results show that the proposed anomaly detection method has good monitoring performance and certain application value.Specifically,the research content of this paper includes the following two aspects:(1)Design and implementation of smart grid anomaly detection algorithm based on graph filter.After the smart grid is constructed as a graph model,the graph high pass filter is designed to monitor the filtered high-frequency component by using the signal data similarity characteristics between each neighbor node in the graph signal.If there is a high-frequency component,it is judged to be abnormal,otherwise it is normal.Then,the abnormal monitoring function of power grid data can be completed by using reasonable judgment method.(2)Further,considering that the grid anomaly data monitoring model with fixed parameters cannot dynamically adjust relevant parameters to meet the more complex requirements of smart grid anomaly monitoring,based on the design of anomaly detection algorithm of graph high pass filter,an anomaly detection model based on graph convolution neural network is further designed to aggregate the neighbor information of each data node in the constructed smart grid model,Thus,it can fully describe the data characteristics of the smart grid,and then complete the anomaly monitoring of the smart grid.According to the given evaluation indicators,it can provide a valuable design reference for practical application.