AUTOMATION OF POWER TRANSFORMER MAINTENANCE THROUGH SUMMARIZATION OF SUBSPACE CLUSTERS

Abstract

Power transformer is considered as critical equipment in power transmission/distribution systems and hence undergoes periodical maintenance for better performance and longer life. The operational condition of a power transformer is continuously monitored by sensing a large number of parameters, which contain hidden patterns indicative of different faulty operational conditions. This paper presents a methodology for automatically identifying such patterns to predict a given faulty condition applying the state-of-art techniques of subspace clustering. The authors propose to summarize an enormously large number of patterns produced by conventional subspace clustering using Similarity connectedness-based Clustering on subspace Clusters (SCoC). The experimentation is done on a real dataset of transformer testing and maintenance records and it is observed that SCoC algorithm proposed by the authors is more effective and efficient in terms of purity and execution time compared to the SUBCLU and PCoC algorithms.

Keywords

• Operational parameters
• Power transformer maintenance
• Preventive maintenance
• Subspace clustering and summarization.