Research And Application Of Distribution Network Fault Prediction

Distribution network fault prediction has always been an important issue in power grid reliability evaluation.Because of its many influencing factors and complex structure,it has been difficult to predict.Appropriate distribution network fault prediction schemes are of great significance for improving the reliability of distribution network distribution,rationally optimizing power grid equipment,saving manpower,and guaranteeing users’ rigid demands in production or life.Based on the analysis of previous research ideas,this paper proposes to apply graph neural network to distribution network fault prediction scenarios and has achieved certain results.First,the graph neural network based on the fixed point theory is used to predict the distribution network fault and analyze the test results.According to the model problems found in the test results,a graph neural network based on convolution is proposed.The physical performance of convolution is to extract the local features of the observation object.In the distribution network,it is also necessary to extract the local features of each target node(neighbor nodes of the target node).In this paper,we first analyze the basic ideas of predecessors applying convolution on the graph and determine the scheme of this paper.Through the experiment,we analyze the performance of each information collection function in the distribution network fault prediction,and select the depth parameter k Value.Tested on the same data set,the experiment shows that the graph neural network based on convolution has achieved better results.Although there is still some distance from the actual application,it has certain reference value for the distribution network fault prediction.