Fault Section Location of High‐Impedance Grounding Faults in Active Low‐Resistance Distribution Networks Based on Multi‐Dimensional Feature Clustering

ABSTRACT

To address the issue that the zero‑sequence current characteristics of single‑phase high‑impedance grounding faults become complex in low‑resistance grounded systems with the integration of inverter‑interfaced distributed generators (IIDGs) and that traditional fault location methods are insufficiently adaptive, a fault section location method based on multi‑dimensional feature clustering of zero‑sequence currents is proposed. First, a zero‑sequence network model of the low‑resistance grounded system with IIDGs is established, and the zero‑sequence current characteristics under different fault positions are analysed. The results show that the zero‑sequence current upstream of the fault point is mainly characterised by low‑frequency resistive components, whereas the current in the region affected by IIDGs may exhibit reverse low‐frequency resistive or high‑frequency oscillatory decaying capacitive characteristics. On the basis of these features, a multi‑dimensional feature set is constructed, integrating time‑domain, frequency‑domain, and improved weighted dynamic time warping (WDTW) similarity information. Non‑metric multidimensional scaling (NMDS) is then employed to reduce the dimensionality of the improved WDTW matrix, and K‐means clustering is subsequently used to achieve automatic fault section location. Simulation results demonstrate that the proposed method can correctly locate the fault section with a transition resistance of up to 4 kΩ, an SNR as low as 10 dB, and sampling frequencies ranging from 4 to 10 kHz. It also remains effective under an overall recording‐device startup delay of 1 ms and inter‐measurement‐point sampling asynchronisation of up to 2 ms.