Statistical analysis of forced outage duration data for subtransmission circuit breakers

Abstract

The main aim of this paper is to study the duration of forced outages for Eskom subtransmission circuit breakers for events recorded over a period of 10 years. Several statistical methods are utilised in the study of the data set. For various variables of events, boxplots were constructed for levels of variables and used to assess the medians, understand the distribution of durations and to gain the understanding of the spreads of durations. Furthermore, for various variables, Kruskal–Wallis tests were conducted to determine, for various variables of the events, whether or not there was statistically significant variation among the levels. Lastly, for each variable, the authors created a multiple linear regression model, with the chosen variable and duration, as a predictor and dependent variables, respectively, and calculating the coefficient of determination calculated for each model, which are then used to rank the variables as effective predictors of duration. This research has shown the power of the utility of techniques from statistics in the analysis of forced outage durations and uncovering useful insights from the data. The insights obtained can have great use in planning asset management strategies with a view to improving the performance of equipment, at large, or subtransmission circuit breakers, specifically.

Keywords

• Boxplots
• Circuit breaker
• Coefficient of determination
• Forced outage
• Kruskal–Wallis test
• Multiple linear regression