Detection of Incipient Fault in Transformer using DGA Based Integrated Intelligent Method

The aim of the work is to develop a reliable composite DGA (Dissolved Gas Analysis) based method to diagnose single and multiple transformer incipient faults. To achieve this objective a two-stage model is proposed. The first stage of the model is the ANN implementation of Dornenburg, Rogers ratio, CEGB, IEC, and Duval triangle methods. ANN-based implementation is carried out to circumvent limitations of the considered methods. Further, to resolve the conflicts of the first stage diagnosis and to predict the most probable single or multiple faults an intelligent rule-based scheme is developed as the second stage of the integrated model. The idea is to exploit the strengths of the different DGA methods to converge to a more reliable diagnostic method using intelligent integrating rules. The proposed method is found to be more reliable and comprehensive in comparison to contemporary methods.