Study On Intelligent Fault Diagnosis Of Power Transformer Based On DGA

The power transformer is an extremely important equipment that undertakes the task of power transmission and transformation in the power system and guarantees the safe and reliable operation of the power system.It has always been a research hotspot in the field of electrical engineering to accurately diagnose its potential faults and reduce the losses caused by faults.Dissolved gas analysis in oil is one of the most widely used methods in transformer fault diagnosis,but traditional fault diagnosis methods based on dissolved gas in oil have some limitations and ambiguities.Therefore,it is necessary to find a more accurate and effective fault diagnosis method.This article combines advanced intelligent methods with traditional methods to diagnose faults in transformers,which has important theoretical and application significance for the healthy and stable development of power systems.The complex working principle of the transformer and the harsh working environment lead to problems such as difficult diagnosis,uncertain results,and unbalanced fault data in traditional transformer fault diagnosis.In response to these problems,this paper improves the non-coding ratio method by in-depth summarizing the mapping relationship between transformer faults and gas in oil,and attempts to construct a transformer fault diagnosis model based on the fusion of rough sets and probabilistic neural networks.Simulation experiments.Experiments show that the rough set fusion improves the diagnostic ability of the probabilistic neural network algorithm,but there is still the problem of unevenly distributed data affecting performance.Therefore,based on the heteroscedastic probability neural network,different parameters are set for different attributes,and an attribute probability neural network based on mixed Gaussian function is proposed.This method uses a mixed Gaussian kernel function to achieve multi-parameter estimation probability density,followed the idea of maximum expected value method,and uses particle swarm optimization algorithm to estimate its parameters.Finally,Lab VIEW is used to design the fault diagnosis software,and the MATLAB probabilistic neural network module is embedded and the visualized operation and extended functions of the fault diagnosis are realized by calling the MATLAB script.The simulation results show that the improved attribute probability neural network algorithm can effectively deal with unbalanced data,improve the accuracy of the algorithm,and has a good diagnosis effect.The design of transformer fault diagnosis system software based on Lab VIEW is conducive to the realization of convenient and efficient transformer fault diagnosis.