Fuzzy information granulated particle swarm optimisation-support vector machine regression for the trend forecasting of dissolved gases in oil-filled transformers

Abstract

In order to achieve accurate trend forecasting of gas contents in oil-immersed transformers, a fuzzy information granulated particle swarm optimisation-support vector machine (PSO-SVM) regression model is proposed in this study. The fuzzy information granulation approach is implemented to transform the original gas data into a sequence of granules, gaining more general view at the data that retains only the most dominant component of the original temporal series. Then a global optimiser, PSO with mutation is employed to optimise the parameters of SVM regression model, avoiding the drawback of premature convergence compared to the standard PSO. Based upon the proposed model, a procedure is put forward to serve as an effective tool for the trend forecasting of transformer gas contents. Results show that this model is capable of forecasting the gas development trend accurately. Moreover, an accurate forecasting interval can provide valuable information for decision making of transformer routine tests or refurbishment.