A Hybrid Model of RVM and PSO for Dissolved Gases Content Forecasting in Transformer Oil
Sheng-wei Fei, Yong He, Xiao-jian Ma and Yu-bin Miao
Affiliation: School of Mechanical Engineering, Donghua University, Shanghai 201620, China.
Keywords: Dissolved gases content, multi-step prediction, relevance vector machine, regression model, time series.
Prediction of dissolved gases content in power transformer oil is very significant to detect incipient failures of
transformer early. A hybrid model of RVM and PSO (PRVM) is applied to predict dissolved gases content in transformer
oil in this paper, and particle swarm optimization is applied to choose the appropriate embedded dimension m because the
choice of the embedded dimension has a great influence on its generalization performance. In this study, traditional support
vector machine is used in comparison with the proposed PRVM method. In order to testify the superiority of PRVM
compared with the traditional support vector machine fully, single-step prediction mode and multi-step prediction mode
are employed respectively. The experimental results indicate that the prediction ability of PRVM is more excellent than
that of SVM in single-step and multi-step prediction. The article also refers some recent patents on a hybrid model of
RVM and PSO.
Rights & PermissionsPrintExport