Fault diagnosis for HVDC converter based on support vector machine

The converter is an important element for the high voltage direct current transmission (High Voltage Direct Current Transmission, HVDC). The fault diagnosis of converters is mostly in the on-off characteristics for diagnosis of fault properties, the lack of diagnosis results of fault location and quantification. This paper presents a HVDC converter fault probability estimation model based on SVM, and establishes the SVM kernel function. The model carries on the probability estimate for the possible faults, using cross-validation method to make sure the SVM parameters well processed, thus overcoming SVM defects of hard-decision outputs. Through the training and testing of the model, the model of fault recognition rate is higher, it has better practicability and popularization.