Predictive learning and information fusion for condition assessment of power transformer

Abstract

To ensure the reliable operation of the power transformer, its conditions must be continuously monitored and assessed. The transformer condition assessment should make use every piece of information (evidence), which includes not only the measurement data of the transformer under investigation, but also the historic data of this transformer and other similar transformers. To acquire an integrated “picture” of transformer health conditions, one needs to combine the diagnosis results obtained from field measurements, laboratory tests, expert experience, utilities practices, and industry standards. This paper applies predictive learning and information fusion techniques for condition assessment of transformer. The predictive learning explores statistical properties from historic data and makes assessment of the property on the transformers. The information fusion integrates various evidences obtained from different sources. This paper develops several predictive learning and information fusion algorithms. Case studies are presented in this paper.