Synchrophasor Assisted Fault Diagnosis using Support Vector Machine

This paper presents a Support Vector Machine (SVM) based fault detection, classifi cation and location using synchrophasor measurements obtained from the optimally placed Phasor Measurement Units (PMUs) for ensuring fault observability. An Integer Linear Programming (ILP) based PMU placement method is proposed, considering the minimization of installation cost as objective with line observability as its constraint. The breaker and half bus-bars scheme is considered at one of the substations to show its impact on the Optimal PMUs Placement (OPP). After the OPP, a SVM based post-fault studies are carried out using the synchrophasor measurements, available from the PMUs. Three types of SVM-Classifi ers (SVM-C) are used for the fault detection, faulted line identifi cation and the fault classifi cation. Further, fault location is carried out using Support Vector Regressor (SVR) in which four SVMs are utilized, one for each fault type. The same classifi cation and regression is carried out using Radial Basis Neural Networks (RBFNNs) and the results obtained from SVM are compared. The performance of the proposed method is studied on WSCC-9 bus system with and without consideration of the breaker and half bus-bar scheme and on New England (NE)-39 bus system.