Mechanical performance degradation modelling and prognosis method of high-voltage circuit breakers considering censored data

Abstract

Due to the untimely deployment of sensors and the early retirement of high-voltage circuit breakers, life-cycle data is missing, leading to an inability to accurately predict the mechanical performance degradation trend. A new modelling and prediction method of mechanical performance degradation of high-voltage circuit breakers considering censored data was proposed. Firstly, multiple imputation by chained equations was used to impute the missing values in the dataset of circuit breaker closing time, creating an interpolated dataset. Secondly, the Nadaraya–Watson kernel regression method was employed to smoothly estimate the interpolated dataset and eliminate data measurement errors. Then, the functional principal component analysis method was utilized to extract the common degradation trend component and deviation component to construct the degradation model. Finally, Bayesian inference was applied to dynamically update the degradation model parameters and predict the degradation trend of the high-voltage circuit breaker. The results showed that the proposed method was capable of achieving better interpolation accuracy under the condition of different closing time censored data of high-voltage circuit breakers. Moreover, as the degradation progressed, the dynamic prediction effect improved. The research can be used to provide an effective decision-making basis for the operation and maintenance strategy of high-voltage circuit breakers.