LS-SVM Forecasting Model for Dissolved Gas in Transformer Oil Using Genetic Algorithm

LS-SVM (least square support vector machines) is widely used in the regression analysis, but the prediction accuracy greatly depends on the parameters selection, in this paper, improved genetic algorithm is applied to optimize the LS-SVM parameters. The improved genetic algorithm is proposed to avoid such phenomenon and advance stability and veracity of the genetic algorithm, and the prediction accuracy is improved Firstly, this paper introduced the principle of LS-SVM and improved genetic algorithm, and gave the optimization parameter flow chart with improved genetic algorithm. Then this algorithm is used to forecast dissolved gas concentration in power transformer oil. Through comparing the forecasting result with the other results, which are forecasted by traditional SVM and LS-SVM, it proved that the method had the higher forecasting precision. Field application showed that the method is effectiveness.