Forecasting of Dissolved Gases in Oil-immersed Transformers Based upon Wavelet LS-SVM Regression and PSO with Mutation

An approach combing wavelet technique with least square support vector machine (LS-SVM) for forecasting of dissolved gases in oil-immersed transformers was proposed in the paper. The algorithm of particle swarm optimization (PSO) with mutation is employed to optimize the hyper-parameters of constructed wavelet LS-SVM regression (W-LSSVR). Evaluation of forecasting performance is based upon the measures of squared correlation coefficient and mean absolute percentage error. On the basis of the proposed approach, experimental results show that the approach is capable of forecasting the dissolved gas contents accurately. In addition, two types of forecasting approaches, the back propagation neural network (BPNN), the radial basis function neural network (RBFNN), are used to compare with W-LSSVR. The comparisons confirm that the W-LSSVR based upon PSO with mutation has a better generalization capability and is more effective than the others. (C) 2016 The Authors. Published by Elsevier Ltd.