Voltage source converter is one of the most important parts of VSC-HVDC system

due to the restriction of the overvoltage and overcurrent conditions

the possibility of breakdown voltage source converter is high and the maintenance is difficult.In order to effectively diagnose the faults of the voltage source converter

the faults characteristics of voltage source converter in VSC-HVDC system are analyzed.An approach to fault classification for voltage source converter is presented in this paper

which based on s-transform and Support Vector Machine.According to the amplitude-frequency characteristics of the DC bus voltage

each frequency components mainly concentrate near the power frequency and distribute intensively in low frequency.The s-transform is used to extract the integer multiple of fundamental frequency component in low frequency.According to the module timefrequency matrixes

the sum of squared modulus and normalized modulus of each frequency component are used as the fault feature vector.The Support Vector Machine with Binary Tree Architecture（SVM-BTA）which parameters are optimized by genetic algorithm is used to accurately classify the fault types

and used the optimized binary tree support vector machine（SVM）to test and train fault signal feature vector sample.The results show that the method can not only reduce the dimension of the feature vector

save the computation time

but also can achieve higher classification accuracy in the case of small sample.