采用改进粒子群优化与电磁时间反演的GIS局部放电定位方法研究

To address large localization errors that arise when idealized parameters are used to construct an equivalent model for conventional electromagnetic time reversal(EMTR)in locating partial discharges(PD)in gas-insulated switchgear(GIS), a high-precision localization method for PD in GIS using improved particle swarm optimization(IPSO)and EMTR is proposed.First, the equivalent model required for EMTR is established to enable time-reversal focusing of electromagnetic waves.Energy accumulation in the spatiotemporal domain is adopted as the focusing criterion instead of the conventional instantaneous peak field strength, substantially enhancing the stability and reliability of focal point localization.Second, energy concentration is defined as the fitness function of the IPSO algorithm；by dynamically updating the virtual-source injection positions, localization errors caused by idealized parameters for the model and deviations in virtual-source injection positions are compensated.Finally, a dynamic inertia weight strategy is introduced so that the particle swarm exhibits stronger global search capability in the early stage and progressively converges toward the neighborhood of optimal solutions in later stages, thus improving overall search performance.Simulation and experimental validations suggest that the proposed method reduces the average localization error by approximately 36% in cylindrical, T-shaped, and L-shaped GIS cavities compared with the conventional EMTR method, significantly improving PD source localization accuracy.