Defect Diagnosis of Disconnector Based on Wireless Communication and Support Vector Machine

Focusing on the shortage of mechanical defect detection and diagnosis technology for disconnectors, a wireless monitoring method for the mechanical state of disconnectors is proposed. The split-core current sensors and improved voltage sensors are used to measure the motor currents and voltages of the disconnector under typical mechanical states at different working voltages. The wireless communication network is used to upload the acquisition data to the cloud server quickly, and the received data are processed by the software system. By comparing and analyzing the curves of current, input power, and output power under different states, it is concluded that the motor output power can adequately reflect the mechanical state of the disconnector. Twenty-three time-domain features of the output power time curve are extracted to form the original feature vector. Kernel principal component analysis (KPCA) method is used to reduce the dimension of the nonlinear features, and the Fisher's criterion function is constructed to determine the width parameter of the kernel function in the feature optimization. Grid search algorithm is used to optimize the kernel parameters of the support vector machine (SVM), and the trained SVM model is used to classify the mechanical state data whose working voltage part is known, and part is unknown, with a classification accuracy of 100%. The results show that the proposed wireless monitoring method can effectively diagnose the mechanical state of the disconnector and has a good generalization ability.