Principal Component and Hierarchical Cluster Analyses as Applied to Transformer Partial Discharge Data With Particular Reference to Transformer Condition Monitoring

This paper analyses partial discharges obtained by remote radiometric measurements from a power transformer with a known internal defect. Since fingerprints of remote radiometric measurements are not available, the formation of clusters with similar features obtained from captured partial discharge data is crucial. Hierarchical cluster analysis technique is used as a method for grouping different signals. Investigation based on Euclidean and Mahalanobis distance measures and Ward and Average linkage algorithms were performed on partial discharge data pre-processed by principal component analysis. As a result of the analysis, a clear separation of partial discharges emanating from the transformer and discharges emanating from its surrounding is achieved; this in turn should enhance the methodologies for condition monitoring of power transformers.

I. Introduction

Power transformer reliability affects both the electric energy availability to the consumer and also the economical operation of the utility. The capital loss of an accidental power transformer outage is often counted in millions of pounds for output loss only, let alone the costs associated with equipment repair or replacement. Because of this economic incentive, preventive tests and on-line monitoring are of benefit to predict incipient fault conditions, and to schedule outage, maintenance and retirement of transformers.