Synchronous Statistical Line Loss Anomaly Diagnosis Method Based On Semi-Supervised Radial Basis Function Neural Network

With the continuous expansion of power grid scale and complex structure,line loss statistics not only has a large workload and low efficiency,but also has an increasing difficulty in abnormal diagnosis of line loss data,which brings new challenges to power grid line loss analysis and development of loss reduction measures.It is urgent to explore new methods to improve the accuracy of abnormal diagnosis of statistical line loss data.Therefore.on the basis of mining abnormal characteristics of statistical line loss,this paper adopts semi-supervised radial basis function neural network method to carry out abnormal diagnosis of statistical line loss,by using this method,the accuracy of abnormal diagnosis of line loss can be effectively improved,and has important theoretical and practical significance for power grid line loss analysis and development of effective loss reduction measures.Firstly,the characteristics of abnormal data of statistical line loss are analyzed in this paper.Based on the current situation of statistical line loss management,this paper analyzes the main causes of abnormal data produced by statistical line loss,and then studies the characteristics and types of abnormal data of statistical line loss.Secondly,the method of extracting abnormal feature of statistical line loss with approximate equal rank sum is studied.Based on the characteristics that the fluctuation amplitude of statistical line loss data differs greatly between normal and abnormal conditions,the fluctuation characteristic index and singular value characteristic index of statistical line loss abnormal data are obtained by using two methods of statistical comparative analysis and singular value decomposition analysis respectively.Based on this,the "rank sum" analysis method was used to extract the "rank sum" difference characteristic index of statistical line loss.Thirdly,the theoretical line loss power corresponding to the statistical line loss is converted into electricity,and the difference characteristics between the corresponding line loss power and the theoretical line loss power in different dimensions are analyzed,and the key factors and mechanism of the difference are deeply analyzed.By using the analysis of "rank sum",the characteristic indexes of "rank sum" difference of line losses over the same period were extracted.Then,the diagnosis method of statistical line loss anomaly based on semi-supervised radial basis function(RBF)neural network is studied.According to the characteristic index of the extracted statistical line loss anomaly data,combining the advantages of semi-supervised learning and radial basis function neural network,this paper proposes a diagnosis method of statistical line loss anomaly based on semi-supervised RBF neural network.By using a large number of unlabeled samples,the training effect of RBF neural network can be greatly improved.It not only has the global optimal performance of traditional RBF neural network,but also can reduce the calculation time while ensuring the diagnostic accuracy.Finally,the actual statistical line loss data of Gansu Hexi Power Grid are used for simulation to verify the feasibility and effectiveness of the semi-supervised radial basis function neural network for abnormal diagnosis of statistical line loss.