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

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.