Online Prediction Method for the Remaining Useful Life of Power Devices Based on Composite Indicator
Xiao Ma, Jianing Wang, Zhaoyang Wei, Lijian Ding
Abstract
Power switching devices are key components of power conversion systems. Their lifetime assessment is critical to the system's safe operation. This study proposes a method for estimating the remaining useful life (RUL) based on the fusion of multiple monitoring indicators to address the problems of low prediction accuracy, low utilization of monitoring indicators, and partial failure information of existing life models. First, the devices’ monitoring quantities of various electro-thermal parameters were obtained through accelerated aging tests. Second, given the limitations of a single indicator for life prediction, a multifactor coupled composite indicator calculation model was derived by combining the Wiener degradation model and mathematical statistics. The results fully considered the influence of various monitoring indicators on the aging of the devices. Then, the probability density function based on the life distribution was used to predict the RUL of the device. The root mean squared error results showed a 5–9 times improvement in the predictive performance of the composite indicator compared with that of a single indicator. Finally, the framework for a real-time prediction system for RUL was proposed. This framework can provide the theoretical basis for the practical application of the proposed method.