Power transformer fault diagnosis based on multi-class multi-kernel learning relevance vector machine

Diagnosis of potential faults concealed inside power transformers is the key of ensuring power system safety. The existing transformer diagnosis methods only infer based on single informative data and it is difficult to detect transformer faults more correctly. In this paper, fault diagnosis based on multi-class multi-kernel learning relevance vector machine (MMKL-RVM) is proposed which can integrate the informative data that can indicate the existence of fault. MMKL-RVM achieves sparsity without the constraint of having a binary class problem and provides probabilistic outputs for class membership instead of the hard binary decisions given by the traditional SVM. Most importantly, MMKL-RVM enables informative integration of possibly heterogeneous informative data or feature spaces in a multitude of ways, from the simple summation of feature expansions to weighted product of kernels. Additionally, Genetic Algorithm (GA) combined with K-fold Cross Validation (K-CV) method is adopted to optimize the kernels parameters in order to enhance the performance of the MMKL-RVM. Experimental results show that MMKL-RVM is capable of more excellent diagnosis accuracy to BP neural network and SVM.