Research On Intelligent Method Of Transformer Fault Diagnosis

Transformer is one of the key equipment in power system with a great responsibility. Its safety and reliability is very important, therefore, the transformer fault diagnosis technology is a very valuable research topic. However, there are many shortcomings in the research of transformer fault diagnosis. In this paper, the advantages and disadvantages of the previous classification algorithms are analyzed,and the new classification algorithm is applied to fault diagnosis of the transformer,which has important theoretical and practical significance. The specific work includes the following parts:1. The structure and mechanism of the transformer are studied, the relationship between the transformer fault and the gas composition in the oil is analyzed, and the advantages and disadvantages of the methods of data cleaning in recent years are analyzed. As for the transformer state data, the fault data is filtered and preprocessed based on the time series method.2. Based on the analysis of the deficiency of the extreme learning machine,the application of the online sequence extreme learning machine in the transformer fault diagnosis model is proposed. The influence of different parameters on the accuracy of the diagnosis model is analyzed by examples and the advantages and disadvantages of two algorithms on the accuracy and efficiency are compared;3. Online sequence extreme learning machine in the fault diagnosis model where the hidden layer parameters are randomly selected is analyzed, so the output is not stable enough, In order to solve this problem, this paper introduces the Particle Swarm Optimization (PSO) algorithm to optimize the online sequence extreme learning machine. Because the PSO is easy to fall into the local maximum,the particle swarm optimization algorithm is firstly optimized, and the optimized particle swarm algorithm is applied to the online sequence extreme learning machine. Online sequence extreme learning machine and PSO are both simulated,and the simulation results show that the proposed method is effective.