Physics-Informed Data-Driven Transmission Line Fault Location Based on Dynamic State Estimation

Abstract

Accurate fault location methods are of great significance for power restoration after system faults. This paper proposes a physics-informed data-driven transmission line fault location method. Two terminal voltage and current sampled value (SV) measurements are typically required. Traditional data-driven methods typically do not consider physical information embedded within the system. Instead, this paper builds the dynamic line model and utilizes the powerful tool of dynamic state estimation (DSE) to track system transients during faults. The fault-related features are extracted via DSE and are utilized as the inputs of the data-driven network to achieve fault location. Numerical experiments in a 500kV AC transmission line system show that the proposed physics-informed data-driven method has higher fault location accuracy in comparison to the traditional data-driven methods without consideration of physics information. The proposed method only needs the fault data window of 5ms after the occurrence of the fault, which is suitable for lines equipped with fast tripping relays. The proposed method is compatible with IEC61850-9-2 standard as it only requires SV measurements with a relatively low sampling rate of 80 samples/cycle. Moreover, although the dynamic line model is utilized for consideration of physics information, the proposed method shows strong robustness against parameter errors.