Partial discharge (PD) represents an early warning sign of insulation degradation in surge arresters—critical devices protecting power systems from lightning and switching transients. Early detection prevents sudden failures, but environmental noise and electromagnetic interference severely distort signals, making precise fault location difficult.
Researchers addressed this challenge by developing a multi-stage signal processing approach. The method first denoises corrupted measurements using three complementary techniques: singular value decomposition removes background noise, adaptive variational mode decomposition (optimized via sparrow search algorithm) isolates frequency components containing discharge pulses, and Teager energy analysis extracts pulse onset characteristics. This combined strategy preserves the fine temporal details of partial discharge events while suppressing interference.
The second innovation enhances time-delay estimation—the core calculation for locating faults along arresters. Rather than conventional cross-correlation, the researchers introduced an HB-weighted quadratic approach that sharpens the correlation peak through dual frequency and amplitude weighting. This weighting concentrates the algorithm's sensitivity on discharge signal characteristics while downweighting noise contributions.
Laboratory validation on a needle-plate discharge platform demonstrated exceptional performance. Across multiple tests spanning distances from under 30 centimeters to beyond 50 centimeters, location errors remained below 0.6%. Simulations achieved 99.9911% detection accuracy—substantially outperforming traditional methods including PHAT-SCOT and NLMS techniques.
The practical significance extends to online arrester monitoring in operational substations. Utilities can deploy this approach for continuous condition assessment, enabling predictive maintenance strategies and preventing unplanned outages. The method's robustness under low signal-to-noise conditions makes it particularly valuable for field deployment where electromagnetic noise is unavoidable. Implementation could allow distributed monitoring networks across transmission and distribution systems, improving grid reliability while reducing inspection costs and equipment replacement cycles.
