Identification Methohd of Grounding Fault for Distribution Station Independent of Zero-sequence Voltage

Fusion identification method is usually used to identify grounding faults by distribution station terminals without access to zero-sequence voltage. It is required that the fusion identification method can adapt to different topologies of distribution stations, and has high identification accuracy for faults with small grounding current. The proposed method combines the physical-model-driven and neural-network methods. First, based on the physical model of grounding faults, this method preprocesses data, selects reference bay and solves the problem of adaptation to different topologies and the number of bays according to the comparison and optimization of zero-sequence current characteristic quantities. And the characteristic quantity of the phase current of a single bay is designed to improve the accuracy of fault identification. Then, a neural network is designed to fuse multiple characteristics to enhance the collaborative judgment ability of different characteristics and reduce the influence of the settings on the sensitivity. Field grounding fault experimental results demonstrate the strong adaptability to different topologies and the high identification accuracy of the the proposed method. © 2021 Automation of Electric Power Systems Press.