Transmission line fault classification based on spatiotemporal characteristic analysis with global and local discriminant analysis

Abstract

To ensure the reliability of power supply, ultra-fast and reliable fault classification of the transmission line is the critical stage of the wide-area transmission network protection. As the distributed generation (DG) current injection to the transmission line, the amplitude characteristic of the break current protection reduces and the system reliability decreases. With the wide-area measurement generated to the power system, an online fault classification method with the global and local discriminant analysis (GLDA) is proposed to enhance the phase characteristic and decrease calculation load. In the single line to ground (SLG) fault process, the local current phase shift is analyzed to enhance the spatiotemporal characteristic. To fully measure the ordering degree of the spatiotemporal characteristic, the axiomatic definition for the characteristic entropy (CE) of the grid impedance in the individual bus in the wide-area transmission network is proposed according to the characteristic distribution. Based on the current spatiotemporal characteristic analysis, a multi-priority GLDA classifies the fault phase in a coordinating way using with the phasor measurement unit (PMU). The multi-priority GLDA-CE fault classification method reduces the computational load, ensures resilience against measurement noise and the transition resistance effect. Compared with the statistical entropy and neural network method, the classification accuracy and response speed of the GLDA-CE fault classification method are validated in the IEEE 39-bus system with 10 PMUs. Compared with the conventional statistical entropy and neural network methods, the GLDA-CE method is able to classify 226 entire power outages and 600 single phase outages within 1.621 s.