Implementation of Extreme Learning Machine to Predict Distribution Power Transformer Lifetime

Abstract

The estimated remaining life of transformers is the most critical factor in maintaining service continuity and quality assurance in distributing electrical energy to customers. Therefore, an accurate approach is needed to predict the remaining life of a distribution power transformer. In this paper, an approach to predict transformer lifetime is carried out by utilizing transformer current, voltage, and temperature. The resulting model estimates the transformer age through Extreme Learning Machine in order to extract the maximum information from the transformer. Tests were performed using the proposed and other classification methods on a 20 kV/380-220V transformer whose remaining life spans between 14 and 26 years. Results show that the proposed method yields the lowest minimum error outperforming the other methods with the selection feature MAE at 0.1076 and MSE at 0.0179. The results further validate that the proposed method can achieve accuracy better than the ones obtained from other transformer remaining service life prediction methods.