基于三维光学指纹和NPSO-KELM的GIL局部放电定位方法

GIL Partial Discharge Localization Method Based on 3D Optical Fingerprint and NPSO-KELM

气体绝缘金属封闭输电线路中局部放电的有效检测及定位对于及时发现绝缘缺陷、提高检修效率至关重要。目前,局部放电光学检测作为一种有效的局放检测方法,具有良好的研究及应用前景。针对GIL气室轴向距离较长、局放源定位困难的问题,提出一种基于三维光信号仿真指纹(three dimensional-optical signal simulation fingerprint,3D-OSSF)和非线性粒子群-核极限学习机(nonlinear particle swarm optimization-kernel extreme learning machine,NPSO-KELM)的定位方法,能够实现局放源的精确定位。该方法将光学仿真数据引入局放源定位中,克服了常规基于指纹的定位方法需要采集大量现场实验数据的难题。通过建立与实验GIL尺寸完全相同的仿真模型,获得不同位置的局放源光学仿真信号,构建包含坐标信息的光学定位仿真指纹库。继而通过NPSO算法对KELM模型进行优化,利用优化得到的NPSO-KELM模型将实测局放光学指纹与指纹库进行模式匹配,得到相应的局放源空间坐标。实验结果表明,该方法的平均定位误差小于lcm,能实现GIL中局放源的精确定位,定位效果明显优于常规KELM算法和BPNN算法。

The effective detection and location of partial discharges in gas-insulated metal-enclosed transmission lines(GIL) is essential for timely detection of insulation defects and improvement of maintenance efficiency. At present, as an effective partial discharge detection method, partial discharge optical detection has good research and application prospects. Aiming at the problem of the long axial distance of GIL air cells and the difficulty of localization of localized source, this paper proposed a localization method based on three dimensional-optical signal simulation fingerprint(3D-OSSF) and nonlinear particle swarm optimization-kernel extreme learning machine(NPSO-KELM). This method can realize the precise positioning of the PD source. This method introduces optical simulation data to the localization of PD source, which overcomes the problem that conventional fingerprint-based positioning methods need to collect a large amount of field test data. By establishing a simulation model with the exact same size as the experimental GIL, the simulated optical signals of the PD sources at different positions were obtained, and an optical positioning simulation fingerprint database is constructed. Then, the KELM model was optimized by the NPSO algorithm, and the optimized NPSO-KELM model was used to match the measured optical fingerprints of the local PD with the fingerprints in the fingerprint database to obtain the spatial coordinates of the PD source. The experimental results show that the average positioning error of this method is less than 1 cm, and it can achieve accurate localization of the localized source of GIL. The positioning effect is significantly better than the conventional KELM algorithm and BPNN algorithm.