A Data-Driven Health Index for High-Voltage SF6 Circuit Breakers

Abstract

High-voltage circuit breakers protect and control the electrical power grid, making their reliability of the essence. Robustly assessing the technical `health' of every circuit breaker in operation is key for an optimal maintenance strategy and safety measures, but is traditionally not performed in a data-driven, statistical fashion. The goal is to design an indexation method for the health of the high-voltage SF6 circuit breakers of TenneT TSO NL using its available (monitoring) data. This indexation is performed through modelling the hazard rate of major and minor component failures with the Cox proportional hazards model. Failure modes and their causes are investigated using large-scale surveys on historical failure data and covariates for the model are chosen accordingly. The hazard rate for minor failures is found to be increasing with the asset's operating voltage level, its average switching frequency, certain manufacturers, and its relative SF6 leakage rate, of which the last relation has not been modelled before to our knowledge. Using the model output, the circuit breakers' probabilities of future failure are estimated, shown to have some predictive power, and incorporated into a color-coded scoring system. Using this quantitative measure of their condition, the TSO should be able to better manage its asset portfolio and risks. However, data quality, especially for major failures, is found to be low and poses a limitation, as well as a possibility for future research.