Prediction of Dissolved Gas Concentration in Oil Based on Fuzzy Time Series

Abstract

The prediction of dissolved gas content in transformer oil is helpful for early detection of latent faults in transformer, and it has important guiding significance for better condition based maintenance. In view of the abundant data of transformer DGA, and that the trend of the change of dissolved gas content in oil under normal running condition is not obvious, a prediction method based on fuzzy time series model is proposed. Consider that the change in dissolved gas content in oil is interaction and influenced, in this paper, the classical fuzzy time series model is improved from the view of domain division, and propose a multi factor fuzzy time series model based on spatial FCM domain partition. The example analysis shows that the method can well fit the changing trend of DGA data, and compared with the classic fuzzy time series model and the one-dimensional FCM fuzzy time series model, the superiority of the improved model in prediction is verified.