A computationally less expensive fault detection technique in VSC-HVDC system using wavelet decomposition and support vector machine classifier

Abstract

The VSC-based HVDC system has gained more popularity due to the advancement in the high power carrying capability of the semiconductor switches. Despite the numerous advantages, the protection scheme faces the challenge to isolate the internal fault within 5 to 6ms. Moreover, the algorithm also deactivates the relay, to operate during the disturbances and external fault. This paper proposes a protection scheme utilizing Discrete Wavelet Transform (DWT) and Support Vector Machine (SVM) classifier. Several types of faults are created on the HVDC transmission lines using PSCAD/EMTDC. The current and voltage signals of the HVDC line are obtained and decomposed using DWT to obtain the detailed coefficients up to the third level. Features obtained from the detailed coefficients are further used for the training of SVM for the detection and classification of the fault. Since the proposed approach uses only the information from one end, it does not rely on communication. To avoid the high computational burden of the Wavelet transform, the protection scheme is designed in such a way that, it performs the DWT, only if the disturbance persists for five consecutive samples.