Method to Detect MMC-HVDC Transmission Line Fault Utilizing Optimal Sample-trained Support Vector Machine
Accurately identifying line fault types in MMC-HVDC systems is of great significance for quickly restoring the normal operation of faulted lines,but high-resistance grounding faults have always been difficult to identify.In this paper,a fault type detection method based on support vector machine is proposed.The method uses empirical modal decomposi-tion to extract a number of high-frequency modal quantities in the fault voltage signal,and uses the particle swarm optimi-sation algorithm to find the optimal weights of each modal quantity and then reconstructs the waveform signal as the opti-mised feature quantity to train the classification model.Simulation results verify that the classification model trained with the optimised samples can accurately identify different types of faults with low sampling frequency and fewer fault wave-form sampling points.
DC transmissionhigh-resistance ground faultempirical mode decompositionsupport vector machinepar-ticle swarm optimisation