Abstract: Pumped storage units are more prone to failure than conventional hydropower units because of their frequent start-up, shutdown and changing working conditions. This paper presents an improved proportional covariate model (PCM) based on a real-time evaluation method for the operational reliability of pumped storage units. This new method first solves for the initial failure rate using an ideal solution approximation method based on the Marxian distance, not demanding large sample failure data; then, a relationship between the failure rate and response covariate is established through a multiple linear regression model to avoid subjective variability brought in by artificial determination of the basic covariate. Finally, operational reliability is updated repeatedly using a multivariate linear function to dynamically reveal the mapping between the condition monitoring data and the reliability. The method has been applied to reliability assessment to the shaft system of pumped storage units, and it achieves satisfactory results that can reflect the operation status of the shaft system in real time, demonstrating its effectiveness and validity.

Key words: pumped storage units, reliability of shaft system, proportional covariate model, the ideal solution approximation method, multiple linear regression