Diagnosis Method Of Transformer Fault Based On Intelligent Extreme Learning Machine

The improvement of transformer fault diagnosis accuracy is conducive to the implementation of equipment operation maintenance and condition maintenance based on state information.Presently,the widely used evaluation method is the three-ratio method,which has incomplete coding,can not cover all fault types,and the evaluation criteria are too absolute.With the input of monitoring equipment and the improvement of detection methods,more and more transformer status data are collected,which provides data source for the research of transformer fault diagnosis method based on artificial intelligent algorithm.These data represented transformer states with different operating conditions,different operating environments,different degrees of error in monitoring equipment,different detection cycles,and different fault reason as well.This paper focuses on how to improve the accuracy of transformer fault diagnosis model based on data from transformer.The main contents of this paper are as follows:(1)Considering the situation that the data formation environment is complex and the way data collected,there may be a small number of outliers.In this paper,the fuzzy C-means clustering algorithm combined with DBSCAN is used to replace the manual data selection method,which realizes the data classification and noises cleaning.The data with the outlier and edge data removed is clearer and improved.It reflects transformer status characteristics clearly.(2)This paper studies learning capacity of transformer fault diagnosis method based on some intelligent algorithm.In order to improve the diagnosis accuracy,a transformer fault diagnosis method based on intelligent extreme learning machine is proposed to transform the data representing transformer state into mathematical model,which puts intuitive and comprehensible diagnosis results.To this end,this paper uses the immune algorithm to improve swarm optimization,which overcomes the shortcomings of the algorithm,and improves the global fast optimization ability.The accuratcy of the model optimized by improved immune swarm optimization is increased.(3)The optimization pursues a high training output accuracy as the target,which leads to the problem of model over-fitting.This paper uses the regular coefficient to adjust the punitive adjustment of the model,which effectively improves the generalization ability of the model.Examples show the effectiveness and feasibility of transformer fault diagnosis method based on intelligent extreme learning machine.Comparing with transformer fault diagnosis model based on extreme learning machine,regular extreme learning machine,optimal regular extreme learning machine,particle swarm optimization extreme learning machine and genetic algorithm optimization extreme learning machine,it obtained a conclusion that the calculating accuracy of the method proposed in this paper is better and the calculation time is shorter as well.