A New Method to Evaluate and Predict the Load Capacity of Transformer in Energy Internet with Data Mining Algorithm

Abstract

Based on realistic transformer dataset, this paper comes up with a method to predict the top oil temperature (TOT) of a main transformer based on the historic TOT, ambient temperature (AT), transformer load (TL) and present AT, TL. Technically, TOT is predicted by striking a balance between univariate time series prediction and multivariate prediction, more specifically, between considering time series features such as trend, seasonality and considering relationship among TOT, AT and TL. From the results, the proposed scheme significantly outperforms the tradition time series model and support vector regression.