Fault section localization for pole-to-ground faults within de-energized areas of DC distribution systems under network parameter uncertainty

Due to limited fault-isolation capability, the initial protection action in DC distribution systems may de-energize a feeder-level or multi-section area containing multiple candidate components. Identifying the specific pole-to-ground (PTG) faulted section within such a de-energized area using recorded fault transients remains challenging under limited measurement coverage and network parameter uncertainty. By exploiting the localized dominant response characteristics of transient zero-mode current (TZMC) during the early stage of PTG faults, this paper proposes a fault section localization method through sparse recovery of the dominant nodal TZMC support. First, the propagation and attenuation characteristics of early-stage TZMC are analyzed to identify a frequency band in which the nodal TZMC responses exhibit localized sparsity. A sparse localization model is then formulated via the impedance mapping between nodal TZMC responses and limited transient zero-mode voltage (TZMV) measurements, accounting for parameter-induced structured mismatch. Furthermore, a topology-aware residual-discriminative matching pursuit (TRMP) algorithm is developed to improve support recovery and thereby enhance fault section localization under such mismatch. The proposed method is validated on a modified IEEE 33-node DC distribution network in PSCAD/EMTDC. Simulation results show that the proposed method can distinguish bus faults from line-section faults and achieves higher localization accuracy than the compared sparse reconstruction methods under varying fault resistances, measurement noise levels, and network parameter uncertainty. Specifically, under network parameter uncertainty, the proposed method achieves average localization accuracies of 96.80 % and 98.99 % for bus-fault and midpoint line-fault scenarios, respectively.