摘要
波形信号是电力设备监测中常见的数据形式,波形信号处理在大数据背景下成为计算和数据双重密集型问题。集合经验模态分解(EEMD)的自适应性在分析非线性、非平稳信号时具有优势,但高计算量限制了其应用。通过对EEMD算法处理波形信号时的并行性分析,在Spark计算平台下设计并实现了波形分段并行与经验模态分解(EMD)过程并行这两种不同结构的并行EEMD算法。波形分段并行适用于较长的波形信号,但结果存在部分误差,而EMD过程并行能获得与原算法一致的结果,但对内存的需求更大,适于数据量不大的波形信号。将并行EEMD算法用于局部放电(PD)波形信号的特征提取,计算固有模态分量的能量参数与样本熵作为特征量。实验结果表明,利用所提特征量能有效区分多种PD类型,所提并行EEMD算法与现有EEMD算法相比计算效率更高,减少了特征提取过程的耗时。
Waveform signals are common data in the condition monitoring of electrical power apparatuses, their processing becomes a data-intensive and computing-intensive problem in the background of big data. Ensemble empirical mode decomposition（EEMD） algorithm, which is adaptive, has advantages in analyzing nonlinear, non-stationary signals, but its application is limited by the high computational complexity. Based on the parallelism analysis of EEMD, two different structure of parallel EEMD algorithm, namely epoch parallel and trial parallel, are designed and implemented on Spark platform. Epoch parallel EEMD, which segments waveform and processes fragments in parallel, is applicable to long waveform, but its results have some errors. Trial parallel EEMD, which parallelizes the empirical mode decomposition（EMD） trials, can get the same results with the original algorithm, but its memory requirement is relatively larger, so it is suitable for the waveform signals with little data volume. The proposed parallel EEMD algorithms are used to extract features from PD waveform signals, and energy parameter and sample entropy of the IMFs are calculated as features. Experimental results show that the features can be used to recognize types of partial discharge. The proposed parallel EEMD algorithms are more efficient than the existing EEMD, which saves the time of feature extraction process.
作者
朱永利
王刘旺
Zhu Yongli;Wang Liuwang(School of Control and Computer Engineering North China Electric Power University Baoding 071003 China;Electric Power Research Institute of State Grid Zhejiang Electric Power Company Hangzhou 310014 China)
出处
《电工技术学报》
EI
CSCD
北大核心
2018年第11期2508-2519,共12页
Transactions of China Electrotechnical Society
基金
国家自然科学基金(51677072)
中央高校基本科研业务费专项资金(2016MS116)资助项目
关键词
集合经验模态分解
并行计算
SPARK
局部放电
波形信号处理
Ensemble empirical mode decomposition
parallel computing
Spark
partial discharge
wave signal processing
