Dynamic State Estimation Enabled Health Indicator for Parametric Fault Detection in Switching Power Converters

This article proposes a parametric fault detection method for switching power converters. The method generates a health indicator to represent the health condition of the entire power converter, and parametric faults are detected with any abnormality of the health indicator. The method first introduces a systematic mathematical modeling framework to describe all the physical laws of the switching power converter during healthy conditions. Afterwards, the dynamic state estimation via batch mode regression is applied to solve the states of the system and to generate the health indicator. The health indicator sensitively reflects the consistency between the actual measurement and the healthy circuit model by taking the statistic characteristics of solution into consideration. The proposed method only requires terminal measurements of the switching power converter, and does not have further assumptions on the topology of the converter. Compared to the existing observer based approaches, the method presents higher sensitivity during parametric faults. Compared to the existing parameter identification based approaches, the method does not need to estimate all the parameters of the converter in order to detect parametric faults. Simulation and experimental results on an example buck converter prove the validity of the proposed parametric fault detection method.