Commutation failure recognition in HVDC systems using wavelet and shannon entropy

Commutation failure (CF) is a serious malfunction in HVDC systems. Fast detection and identification is important to avoid HVDC system block or even further deterioration of the whole system. This paper proposes a novel approach of CF recognition using wavelet transform and Shannon entropy technique. Only phase A voltage signal at inverter is used as the input of the new method. AC voltage is decomposed by discrete wavelet transform to get details and approximations on all scales. High frequency details are used to locate the fault time. A fault entropy feature matrix T is formed for fault classification. CF can be identified by comparing the Euclidean distances of its feature vector to those of the known fault types in matrix T. The proposed approach is verified by fault signals simulated in a complete 12-pulse HVDC transmission system in Matlab/Simulink. The results have approved its feasibility and preciseness.