Research On Support Vector Machine-Based HVDC System Fault Diagnosis

HVDC is a new type of power transmission technology developing rapidly in recent years, mainly applied to faraway and high capacity power transmission, power system networking and submarine or underground cable power transmission. System faults will result in production pause or confusion, so, fault diagnosis is the core technology of HVDC and also the precondition of system protection. Considering the existing state that researches on HVDC system fault diagnosis of home and abroad are relatively rare, types and features of various HVDC system faults are analyzed in this paper; S transform and Support Vector Machine (SVM) are used in system fault diagnosis and a few corresponding key technology problems are deeply investigated on the basis of simulation.First, some traditional fault diagnosis methods and signal processing methods are analyzed, and then S transform and SVM are respectively applied to feature extraction and fault classification, by comparing the merits and drawbacks of various methods. S transform is a kind of reversible time frequency local analysis method that can be used to extract time frequency features of signals and solve the scale parameter selection problem in wavelet transform. SVM is a novel learning method based on statistic learning theory and has been a hot issue in machine learning area in solving small sample, nonlinear and high dimensional pattern problems.After analyzing the common faults and features of HVDC, an electro-magnetic transient state simulation software named EMTDC/PSCAD is used to set up a simulation model of HVDC system to investigate the typical power system faults. Features of waves are analyzed and DC voltage & current signals which can reflect fault features are selected to be S transformed and then fault samples composed of effective features are extracted. SVM training and classification testing are then...