Mechanical Fault Diagnosis of Transformer On-Load Tap-Changer Based on Improved Variational Mode Decomposition and Support Vector Machine

Transformer on-load tap-changer (OLTC) is a key component of transformer. Vibration detection method can be applied to the mechanical condition monitoring and fault diagnosis owning to its non-invasive and high-sensitive property. However, several problems such as modal aliasing and feature one-sidedness exist in the current signal processing and fault diagnosis method for OLTC vibration signal. In the paper, an OLTC simulation experiment platform and a vibration signal sensor system are built, and three typical defects, including spring elasticity change, contact wear and contact looseness are set. The vibration signals collected by accelerometers are decomposed into a series of modal components with narrow band and distinguishing center frequency by improved variational mode decomposition (VMD). Then the features including energy, permutation entropy and singular values of each modal component are extracted, and Laplacian-score is used to select the features with better discriminative ability. Finally, the support vector machine (SVM) optimized by genetic algorithm (GA) is applied to fault diagnosis. The result shows that the proposed feature extraction method provides comprehensive and distinguishing features in each operation state, and the proposed fault diagnosis method has higher identification accuracy compared with conventional method.