A novel setting free approach to differentiate fault and power swing using support vector machine

Undesired tripping may occur in protection relays due to the power swings leading to shutting down of power utility equipment. Conventionally, the rate of speed of the impedance locus measured by the relay is utilised to differentiate the fault and power swing. Nowadays, the inertia of the system is lowered due to the proliferation of distributed generators and therefore it is very difficult to adopt a threshold limit to distinguish the fault from power swing. A setting or threshold free approach based on support vector machine to detect power swing is proposed in this work. The prominent SVM feature like load angle calculated indirectly from relay impedance is a novel way adopted in this scheme. The remaining SVM features selected are also a good combination of statistical and electrical parameters. Results also show that the scheme is capable to identify both symmetrical and asymmetrical faults during power swing. All the simulated case studies are also tested in a transmission line prototype set-up in the laboratory.