Remaining Useful Life Prediction of Thyristors for UHVDC Converter Valve Based on Bivariate Nonlinear Wiener Process

ABSTRACT Remaining useful life (RUL) prediction is a key technology for prognostics and health management (PHM). However, the conventional thyristor RUL prediction method that solely relies on a single degradation indicator proves insufficient in comprehensively reflecting the thyristor's overall health status. In this paper, we proposed a thyristor RUL prediction method based on the bivariate nonlinear Wiener process. First, a thyristor degradation model with a single degradation indicator is established based on the nonlinear Wiener process. The model can describe the nonlinearity and stochasticity of the thyristor degradation process. Then a two‐performance‐dependent thyristor degradation model is established based on the Copula function. Finally, the Markov Chain‐Monte Carlo (MCMC) method is adopted to estimate the unknown parameters of the model. The on‐state voltage and reverse recovery charge serve as key degradation indicators, and the proposed method is validated by relying on the accelerated life test data, comparing the RUL prediction results of different models. The results show that the proposed method can more comprehensively describe the health state of the thyristor, and the result of RUL based on the proposed model is closer to the actual results.