A HVDC System Fault Diagnosis Method Based on PSO-LightGBM Algorithm

Due to the complexity of their structure and the particularity of their application environments, aircraft High-Voltage Direct Current (HVDC) systems are prone to faults, with inter-module failures complicating fault diagnosis. To address this issue, a fault diagnosis method for HVDC systems has been developed. Feature extraction methods were designed for the rectifier, BUCK converter, and inverter, respectively, with the sum-to-amplitude ratio of signals selected as a feature for the rectifier; multi-scale skewness was proposed for the BUCK converter; and the ratio of the signal's average to peak absolute value was chosen for the inverter. Subsequently, the PSO-LightGBM algorithm was proposed, which employs the LightGBM algorithm for classification and utilizes a particle swarm algorithm to optimize the parameters of the LightGBM, establishing the optimal model. The experimental results demonstrate that the proposed method can accurately achieve fault diagnosis in HVDC systems.