A Feature Extraction Method for Soft Fault Diagnosis of DC/DC Converters Using Wavelet Analysis and Gramian Angular Field

DC/DC converter is the key component of power supply system, its safety and reliability ensure the stable operation of the system. Under various environmental stresses, the parameters of the converter gradually deviate from their initial values as the running time increases. When the parameters exceed a certain threshold, the system will suffer from soft faults. This paper presents a novel feature extraction method for soft fault diagnosis of DC/DC converters using wavelet analysis (WA) and gramian angular field (GAF). The degradation index is the ripple of output voltage, After the correlation time-frequency features are extracted by WA, the correlation features are cross-fused by GAF method. On the basis of fusion features, K-nearest neighbor (KNN) network is established for soft fault detection and location. Since no additional hardware or signal injection is required, the entire application process is low in cost and complexity. Simulation and experimental results show the proposed method has a good diagnosis effect on various soft faults.