Fault Diagnosis of a Transformer using Fuzzy Model and Unsupervised Learning

Abstract

In this paper a power transformer fault diagnosis system (PDFDS) based on soft computing and computational intelligence is proposed. Fault diagnosis and analysis is an integral part of operational reliability. Systems like SCADA collect data of various equipment in power system network, however, fails to provide a critique fault diagnosis for the same which further leads to additional cost of replacing the equipment. This paper proposes a supervised-unsupervised predictive model for the data collected from various power transformers across Himachal Pradesh and IEC 10 database. To identify the different fault types in a transformer a fuzzy model is developed using the DGA interpretation techniques. Since, not all data samples in the collected dataset fall under the standards specified in the ratio tables it thus becomes difficult to identify the type of fault for such cases. To overcome this an improved fuzzy model with unsupervised clustering algorithm or Fuzzy Clustering means is used. Employing this improved model optimizes the data before feeding it to the different predictive machine learning models. Further, a particle swarm optimization algorithm with passive congregation is employed to optimize the performance of these machine learning models.