Statistical Analysis of Power Cable Failures and Maintenance Optimization

Abstract

In many parts of the world the distributed cables are ageing and many of them have surpassed their design life, whilst in developing countries such as China an increasing amount of cables have been recently installed in their respective power network where manufacturing and installation deformities are causing problems. The local degradation in a small section of a cable circuit and global degradation in the entire cable are responsible for random and ageing failures, respectively. The risk of failure translates into the financial burden for both utility and its customers. The utility constantly faces the stress of minimising the risk of unplanned outages. This necessitates the need for the development of a methodology to strategically minimise the risk of failure by proactive maintenance. The immediate objective of this research was to develop a methodology to estimate the cable expected life and predict the failures to propose a cost-effective proactive maintenance strategy.In this research, a stochastic electro-thermal model was developed to estimate the expected life of the cable and predict the failures due to global degradation or ageing in insulation; and a non-homogenous Poisson process model was developed to predict the failures due to all causes. Both the models were integrated to predict the annual number of failures due to ageing, random and the combined effect of both causes. The integration of both the models enabled the development of a maintenance optimization model which helped in achieving the optimal decision at each stage of the cable service life. The research has also shown the potential of mixed categorical and numerical data in understanding the failure behaviour and failure trend. A visualization method based on Multiple Correspondence Analysis was developed to enrich the view and understanding of cable failure behaviour by allowing the user to visualize the preliminary pattern and associations which get obscured in high-dimensional multiple variable data. In future, advanced statistical models and data analysis techniques will be developed to unleash the potential of the large or big data available to the utility.

Date of Award	2016
Original language	English
Awarding Institution	Glasgow Caledonian University
Supervisor	Chengke Zhou (Supervisor), Geraint Bevan (Supervisor) & Babakalli Alkali (Supervisor)