基于UHF信号特征参量优化提取的GIS多局放源分离技术研究

Gas insulated system (GIS) is widely used in power system since its unique advantages. However, the insulation breakdown in GIS is still the practical and difficult problem in power system，especially a few serious breakdown accidents of 550kV and 1100kV GIS happened in China in recent years. The detection and diagnosis of various insulation defects in GIS accurately are the hot issue and difficulty in the currently research field. The condition assessment technology of GIS is based on the precondition of single insulation defect. However, several insulation defects sometimes may exist in GIS actually. The coupling influences of signals excited by multiple partial discharge (PD) sources are formed in this condition, which brings the difficulty of detection and assessment of GIS condition. Focused on this problem, the simulation and experimental model of multiple PD sources is built in this research issue. The propagation and coupling characteristics of the UHF signals excited by multiple PD sources are discussed, and the evolution relationship between UHF signals and development of multiple PD is analyzed. The separation feature of multiple PD UHF signals is extracted optimization. By combined the multiple intelligent algorithms together, the cluster and pattern recognition methods are proposed to separate and indentify the types and each development stage of multiple PD sources. This research is hopeful to promote the multiple PD sources diagnosis and application in field.

气体绝缘组合电器(GIS)因其突出优势被广泛应用于电力系统中，但其绝缘的放电击穿问题一直是实际难题，特别是我国近年来出现了多起550kV和1100kV GIS重大事故。如何准确检测与诊断GIS的绝缘缺陷，是国内外关注的研究热点和难点。目前GIS状态评估技术主要针对单绝缘缺陷，而实际情况是GIS内部有时会存在多个绝缘缺陷，多个局部放电源信号之间互相耦合影响，这给状态检测与评估带来了难度。本项目针对这一问题，通过建立GIS中多局放源仿真与实验模型，探讨多局放源UHF信号传播与耦合特性，分析多局放源发展过程与UHF信号间的演化关系，优化提取基于UHF信号的GIS多局放源的分离表征参量，提出多种智能算法联合的聚类和模式识别方法，分离和识别GIS多局放源的类型和各自发展阶段，以期推进GIS多局放源的诊断分析研究与现场应用。

气体绝缘组合电器(GIS)因其突出优势被广泛应用于电力系统中，但其绝缘的放电击穿问题一直是实际难题，目前GIS状态评估技术主要针对单绝缘缺陷，而实际情况是GIS内部有时会存在多个绝缘缺陷，多源信号之间互相耦合影响，这给状态检测与评估带来了难度。本项目针对这一问题，分别针对GIS内多源UHF信号的仿真、实验以及分离识别等方面开展了相关研究，首先采用电磁波仿真软件，建立了与实验平台一致的252kV GIS多局放源UHF信号仿真模型，分析研究了GIS UHF信号模式特征以及传播与接收特性.同时在已有的252kV GIS实验平台上，结合脉冲电流法和UHF法，开展了GIS多局放源下放电演化和发展特性研究。通过建立多局放源UHF信号检测数据库，分析了GIS多局放发展过程与UHF信号间的关系，并与仿真结果进行比较。此外也分析GIS UHF信号定位的关键参数，确定了GIS多源局放下的定位算法。最后基于GIS多源局放UHF信号检测数据库，以放电源定位和信号特征参量为基础，分析GIS设备多源局放各发展阶段下UHF信号特征。此外，本项目提出了以最优分离效果为目标的特征参量优化提取算法，建立了多智能算法联合的聚类和模式识别模型，用以分离和识别多源局放的类型和各自发展阶段。本项目的研究成果已应用于莲都1100kV GIS、洪明变550kV GIS等多个现场案例中，取得了良好的应用结果，具有广阔的应用前景。