High-voltage circuit breakers (HVCBs) play a substantial protection role in power networks. The reliable operation of these critical components leads to an increment of resiliency and safety of power systems. It is essential to design a fault diagnostic system that detects the defects in preliminary levels and identify the origins to establish a precise maintenance task. This paper focuses on coil current and contact travel waveforms as significant signals that bear helpful information about the fault occurrence for a typical EDF, 72.5 kV, SF6 HVCB. Healthy and faulty signals simulated based on Michael Stanek's HVCB model in MATLAB, with performing some modifications in the actuating coil and operating mechanism. In the first step, to arrange an efficient fault recognition system, neural network and support vector machine (SVM) have been designed using the information of 475 simulated healthy and faulty HVCBs and verified for 200 new samples. In the second step, to improve the classification results, an additional distinction algorithm has been recommended for the cases in which two failure modes are detected by the classifier. Since any failure mode's impact on the selected features is different, the proposed diagnostic method makes a decision, between two classes of faults, based on the extracted pattern of each failure mode. The recommended method, which is a combination of commonly used classification techniques and the defined algorithm, leads to the more accurate diagnosis.
Read full abstract