Data mining techniques for transformer failure prediction model: A systematic literature review

Abstract

Transformer failure may occur in terms of tripping, resulting in an unplanned or unseen failure. Therefore, a good maintenance strategy is an essential component of a power system to prevent unanticipated failures. Routine preventive maintenance programs have traditionally been used in combination with regular tests. However, in recent years, predictive maintenance has become prevalent due to the demanding industrial needs. Due to the increased requirement, utilities are persistently looking for ways to overcome the challenge of power transformer failures. One of the most popular ways for fault prediction is data mining. Data mining techniques can be applied in transformer failure prediction to provide the possibility of failure occurrence. Thus, this study aims to identify the common data mining techniques and algorithms that are implemented in studies related to various transformer failure types. The accuracy of each algorithm is also studied in this paper. A systematic literature review is carried out by identifying 160 articles from four main databases of which 6 articles are chosen in the end. This review found that the most common prediction technique used is classification. Among the classification algorithms, ANN is the prominent algorithm adopted by most of the researchers which has provided the highest accuracy compared to other algorithms. Further research can be done to investigate more on the transformer failures types and fair comparison between multiple algorithms in order to get more precise performance measurement. © 2019 IEEE.