Directional, High-Impedance Fault Detection in Isolated Neutral Distribution Grids

This paper presents an algorithm for fast, directional detection of phase-to-ground faults in isolated neutral distribution grids. The method reacts directionally and very fast to fault conditions (1-2 samples after fault inception, even without large di0/dt). Moreover, the method is able to detect low- and high-impedance faults (HIF), permanent or transient faults. The algorithm is independent of the sampling frequency, very robust to noise, and does not need calibration of the sensors. The direction detection method relies on the fault model identification, based on the zero-sequence grid capacitance estimate for isolated neutral grids. The estimate is obtained from the fault circuit differential equation, solved with the least-squares method, where appropriate numerical modifications have been introduced. The sign of the estimate provides the directional information, whereas the magnitude of the modified estimate indicates the fault. The directional principle is validated in simulation and the fault detection method is validated with real fault records from the RedNA project. The paper reworks the equations that define the nondetection zone, so that the method is comparable to the classical residual overcurrent and overvoltage functions, typically used to detect HIF.