Method for Evaluating the Health Status of Key Component Within Power Converter System Based on Particle Swarm Optimization Algorithm

To address the challenge of small-sample fault data in circuit breakers within power converter systems, this study employs the Relevance Vector Machine (RVM) method for health status assessment of circuit breakers. To overcome the dependency of RVM’s recursive parameter estimation on initial values, the Particle Swarm Optimization (PSO) algorithm is utilized for global search and automatic optimization of hyperparameters. Firstly, an RVM-based training model is established. Secondly, the PSO algorithm is applied to optimize the initialization of key RVM hyperparameters (e.g., kernel function parameters, noise variance). Subsequently, the optimized parameters are used for model training and status assessment. Finally, the effectiveness and robustness of the proposed method are validated using degradation data from power converter circuit breakers.