利用累积能量函数特征参量优化提取的多源局部放电信号分离技术
发布时间：2019-02-16 07:14
【摘要】：变电设备内部有时会存在多个缺陷的局部放电,其放电模式识别及危险度评估的难度大大增加,为更有效地诊断设备绝缘状况,该文提出了一种基于累积能量函数特征参量优化提取的多源局放分离技术。利用时频域累积能量函数表征脉冲电流脉冲或特高频(UHF)信号的时频域变化,并采用数学形态学梯度运算提取了时频域累积能量的上升陡度参量。提出了以上升陡度参量的标准差作为分离性能评价指标,优化选取数学形态学梯度运算中的结构元素长度,提取此时的上升陡度参量,达到最优分离效果的目标。最后在实验室252k V GIS模型内建立了3种典型多缺陷模型,将所提出的多源放电分离技术应用于该混合缺陷放电UHF信号的分离,进而将该方法成功应用于一起现场1100k V GIS多源局放案例。结果表明,特征参量优化提取分离方法适用于内外置UHF传感器信号,在多源放电混合UHF信号分离中具有良好的应用效果。
[Abstract]:In order to diagnose the insulation condition of the equipment more effectively, there are some partial discharges which have many defects in the transformer equipment, so it is more difficult to identify the discharge mode and evaluate the risk degree. In this paper, a multi-source partial discharge separation technique based on the feature parameter extraction of cumulative energy function is proposed. The time-frequency domain variation of pulse current pulse or ultra-high frequency (UHF) (UHF) signal is characterized by time-frequency cumulative energy function, and the ascending steepness parameter of time-frequency cumulative energy is extracted by mathematical morphological gradient operation. The standard deviation of ascending steepness parameter is used as the evaluation index of separation performance. The length of structural elements in mathematical morphological gradient operation is optimized and the ascending steepness parameter is extracted to achieve the goal of optimal separation effect. Finally, three typical multi-defect models are established in the 252kV GIS model of laboratory. The proposed multi-source discharge separation technique is applied to the separation of UHF signals from the mixed defect discharge. Furthermore, the method is successfully applied to a field case of 1100kV GIS multi-source PD. The results show that the method of extracting and separating characteristic parameters is suitable for the internal and external UHF sensor signals, and has a good application effect in the multi-source discharge mixed UHF signal separation.
【作者单位】： 国网浙江省电力公司电力科学研究院;国网浙江省电力公司;电力设备电气绝缘国家重点实验室(西安交通大学);
【基金】：中国博士后科学基金资助项目(2015M580848) 国家自然科学基金项目(51607140) 国网浙江省电力公司科技项目(5211DS15002P)~~
【分类号】：TM855
[Abstract]:In order to diagnose the insulation condition of the equipment more effectively, there are some partial discharges which have many defects in the transformer equipment, so it is more difficult to identify the discharge mode and evaluate the risk degree. In this paper, a multi-source partial discharge separation technique based on the feature parameter extraction of cumulative energy function is proposed. The time-frequency domain variation of pulse current pulse or ultra-high frequency (UHF) (UHF) signal is characterized by time-frequency cumulative energy function, and the ascending steepness parameter of time-frequency cumulative energy is extracted by mathematical morphological gradient operation. The standard deviation of ascending steepness parameter is used as the evaluation index of separation performance. The length of structural elements in mathematical morphological gradient operation is optimized and the ascending steepness parameter is extracted to achieve the goal of optimal separation effect. Finally, three typical multi-defect models are established in the 252kV GIS model of laboratory. The proposed multi-source discharge separation technique is applied to the separation of UHF signals from the mixed defect discharge. Furthermore, the method is successfully applied to a field case of 1100kV GIS multi-source PD. The results show that the method of extracting and separating characteristic parameters is suitable for the internal and external UHF sensor signals, and has a good application effect in the multi-source discharge mixed UHF signal separation.
【作者单位】： 国网浙江省电力公司电力科学研究院;国网浙江省电力公司;电力设备电气绝缘国家重点实验室(西安交通大学);
【基金】：中国博士后科学基金资助项目(2015M580848) 国家自然科学基金项目(51607140) 国网浙江省电力公司科技项目(5211DS15002P)~~
【分类号】：TM855
【参考文献】
相关期刊论文 前5条
1 黄亮;唐炬;凌超;张晓星;;基于多特征信息融合技术的局部放电模式识别研究[J];高电压技术;2015年03期
2 司良奇;钱勇;白万建;叶海峰;胡岳;盛戈v，
本文编号：2424186
本文链接：https://www.wllwen.com/kejilunwen/dianlidianqilunwen/2424186.html
