云平台下并行总体经验模态分解局部放电信号去噪方法  

Research on Parallel Ensemble Empirical Mode Decomposition Denoising Method for Partial Discharge Signals Based on Cloud Platform

作　　者：宋亚奇[1,2] 周国亮[1,2] 朱永利[1,2] 李莉[1,2] 王德文[1] 

机构地区：[1]华北电力大学控制与计算机工程学院,保定071003 [2]华北电力大学新能源电力系统国家重点实验室,北京102206

出　　处：《电工技术学报》2015年第18期213-222,共10页Transactions of China Electrotechnical Society

基　　金：国家自然科学基金(61074078);中央高校基本科研业务费专项资金(13MS88;13XS30);新能源电力系统国家重点实验室和河北省自然科学基金(F2014502069)资助项目

摘　　要：信号去噪是对输变电设备进行在线监测和诊断时首要解决的问题。鉴于总体经验模态分解(EEMD)方法对局部放电信号进行去噪的优势,设计了基于Map Reduce模型的并行化EEMD算法(MR-EEMD),利用云平台提高算法的计算效率。在对分段包络线进行重构时,针对矩形窗的固有缺陷,提出了基于局部平稳度的自适应分段包络线重构算法(LF-ASER)进行分段边界的补偿处理,使重构的包络线误差减小到给定阈值范围内。实验结果表明MR-EEMD算法相对于EEMD性能提升显著,适合处理变压器的局部放电等高采样率信号,同时保持了EEMD去噪效果,并获得较高的可扩展性和加速比。Signal denoising is the primary issue when conducting online monitoring and diagnosing of electric transmission and transformation equipments. In view of the advantage of ensemble empirical mode decomposition（EEMD） for partial discharge signal denoising, the parallel EEMD algorithm based on Map Reduce model, named MR-EEMD, is designed to improve the computational efficiency by taking advantage of the cloud platform. In consideration of the inherent defects of the rectangular window, the local flatness-adaptive segmentation envelope reconstruction algorithm（LF-ASER） is proposed to compensate segmented boundary so that the envelope error can be reduced to a given threshold range. The experimental results show that MR-EEMD can be executed much faster than EEMD for the transformer partial discharge high sampling rate signal and maintains good denoising results, high scalability, and speedup.

关 键 词：局部放电 信号去噪 总体经验模态分解 MAPREDUCE 包络线重构 

分 类 号：TN911[电子电信—通信与信息系统]