A Method of Kernel Principal Component Analysis and Machine Learning Algorithms for Fault Diagnosis of Power Transformers

Abstract

As one of the most important equipment in the power system, it is of great significance to conduct fault diagnosis research on transformers. Aiming at the problem of difficult selection of parameters for support vector machine (SVM) in transformer fault diagnosis, a fault diagnosis model based on the improved pelican optimization algorithm (IPOA) optimized SVM is proposed. First, the standard pelican optimization algorithm (POA) is enhanced by introducing the Tent chaotic mapping, Levy flight strategy, and adaptive weighting strategy, and the superiority of the IPOA algorithm is verified by comparing its performance with other intelligent algorithms across four test functions. Second, feature dimensionality reduction of the data samples is performed using kernel principal component analysis (KPCA), and the IPOA algorithm is used to optimize the SVM parameters and then establish the transformer fault diagnosis model based on IPOA-SVM. Finally, the POA-SVM, northern goshawk optimization (NGO)-SVM, GWO-SVM, whale optimization algorithm (WOA)-SVM, and particle swarm optimization (PSO)-SVM models are used to conduct comparative experiments with the proposed method. The results show that the diagnostic accuracy of the IPOA-SVM model reaches 95%, which is 3.75%, 5.0%, 6.25%, 7.5%, and 8.75% higher than that of the POA-SVM, NGO-SVM, GWO-SVM, WOA-SVM, and PSO-SVM diagnostic models, respectively; and the proposed model exhibits better stability and greater adaptability.