Research on IGOA-LSSVM based fault diagnosis of power transformers

Abstract

Power transformer is an important part of power equipment, and its functionality affects the proper operation of the whole power network. In order to diagnose power transformer faults effectively, the authors propose a fault diagnosis strategy based on an improved locust optimization algorithm for least squares vector machines (IGOA-LSSVM). Firstly, it was required to address the problem that the diagnostic prediction accuracy of the least squares vector machine is reduced due to its parameters. So this paper introduces the locust optimization algorithm with simple algorithm structure and good performance for optimizing the parameters. And at the same time, the authors generate an improved locust optimization algorithm with self-learning factors, proportional weight coefficients and Levy flight strategy. Secondly, the improved locust optimization algorithm is used for optimizing the least squares vector machine parameters. Finally, in the simulation experiments, the results of the benchmark test function illustrate that the IGOA algorithm has better performance, and the test results of a fault samples diagnosis of the power transformer equipment illustrate that the IGOA-LSSVM has good prediction effect and improves the fault identification accuracy compared with ACO-LSSVM and PSO-LSSVM in five types of fault diagnosis.