A Novel Condition Assessment System for Underground Distribution Applications

This paper discusses an automated measurement- based condition assessment system for underground distribution. The core components of the system include rule-based and SVM-based pattern classifiers, which utilize novel features in the time and wavelet domains to classify underground cable patterns successfully into various classes based on a measure of severity. The performance of the developed system was tested and evaluated using field data from an underground distribution system. The satisfactory classification results and applications for condition assessment are presented and discussed. The results demonstrate the ability to pick-up and follow the progressive deterioration of the underground system towards an imminent fault by monitoring its incipient behavior over time.

I. Introduction

Underground distribution systems are composed of many assets that should work together to deliver reliable power to end customers at lower voltages. Many of the system components—particularly underground cables—are subject to deterioration and eventual failure due to a number of aging factors including electrical, mechanical, and environmental stresses [1]. To maintain these assets, electric utilities typically perform scheduled inspections and regular maintenance work as a general practice. However, the need for reliability improvement and cost optimization in this area have made it necessary to shift from calendar-based to condition-based maintenance. This maintenance strategy requires online or offline assessment of the condition of the assets to determine the appropriate maintenance work.