Online Health Status Assessment of Metro Auxiliary Inverters Based on an Improved D-S Evidence Theory

Inverters are widely applied in aviation, distributed power grids, and vehicles, where their health status directly impacts the stable operation of entire systems. Existing health assessment methods suffer from poor real-time performance, require additional measurement circuits, and are prone to misjudgment, while failing to adequately address slow degradation behaviors during inverter operation. To address these challenges, this study proposes an inverter health assessment method based on an improved D-S evidence theory. First, based on the practical requirements of subway auxiliary inverters, 13 key evaluation indicators were selected. Subjective weights were obtained using the Analytic Hierarchy Process (AHP), while objective weights were derived through the Critic method, credibility, and falsity weighting. These were then fused using game theory to obtain composite weights. Next, after data normalization, a ridge-type membership function was employed to describe health state uncertainty. Finally, the improved D-S evidence theory integrates multi-source information to achieve online health status assessment. Experimental validation demonstrates that this method effectively evaluates the impact of IGBT failures, sensor malfunctions, and capacitor–inductor degradation on the inverter. It exhibits strong robustness under DC voltage fluctuations and load variations, enabling real-time output of health scores and grades to provide a reliable basis for maintenance decisions.