Reliability of Power Transformers at Kiira and Nalubaale Power Stations

Abstract

This research was a case-study aimed at investigating and developing a statistical transformer model able to predict future failures at Nalubaale and Kiira power stations. The Weibull distribution was identified as an appropriate statistical technique that was developed into a lifetime mathematical model. Quantitative historical data of transformer installation and failure times were collected and exploited to determine the desired lifetimes while the Maximum Likelihood Estimator allowed appropriate data fitting and provision of the most likely parameter estimates. The Rank Regression method and the Weibull log-plotting approaches were also investigated for comparison purposes and no significant departures were noted in the results between the methods. Probabilities of transformer failure were determined using the developed Weibull models. Power transformers are expensive equipment and are critical components of an electric grid system. In-service failure is usually catastrophic to the grid reliability and consequential financial loss to the utility. The majority of transformers at these power stations under study, are less than 20 years old with a few exceeding 50 years of age. As part of Asset management, a transformer retirement policy plan is required at these stations which is the subject of the study so that transformers are proactively replaced at the opportune time to avoid in-service failure and or pre- mature replacement. The study shows that presently, over 85% of the transformers under study are in excellent condition with very low hazard rates and the mean age before failure is 46.04 years. This was used to predict the number of transformers that were likely to fail by that time. However this fell short of adequately addressing the research question and areas requiring further adjustment were noted. The transformer population size at these plants is limited thus data stratification of any type is not necessary in any lifetime study to avoid small sample size problems. The study further shows the old transformers that are 50 years and above should be considered for retirement when an opportunity arises because they are under-going ageing with very high probability of failure to avoid in service failure. All other specific objectives were achieved.