计算机工程与应用 ›› 2017, Vol. 53 ›› Issue (7): 256-262.DOI: 10.3778/j.issn.1002-8331.1503-0211

• 工程与应用 • 上一篇    下一篇

极限学习机延拓的BS-EMD端点效应抑制算法及应用

郭  瑞,樊亚敏   

  1. 辽宁工程技术大学 电气与控制工程学院,辽宁 葫芦岛 125105
  • 出版日期:2017-04-01 发布日期:2017-04-01

Algorithm based on extreme learning machine to restrain the end effect of BS-EMD and its application

GUO Rui, FAN Yamin   

  1. School of Electrical and Control Engineering, Liaoning Technical University, Huludao, Liaoning 125105, China
  • Online:2017-04-01 Published:2017-04-01

摘要: 针对希尔伯特-黄变换过程中经验模态分解出现的端点效应问题,采用极限学习机(Extreme Learning Machine,ELM)算法对原始数据序列分别向左右两端延拓,对扩展后的数据序列用B样条插值函数求其平均曲线,在此基础上进行下一步分解,结束分解后摒弃两端延展的数据,使算法得到优化,起到了抑制端点效应的作用。通过与未经延拓,BP神经网络延拓和支持向量机延拓各项指标的对比分析表明,该算法不仅有效抑制了经验模态分解过程中的端点效应,在预测速度和分解精度上都有一定的优势。将该方法应用于电力系统的谐波分析中,仿真结果表明该方法能有效抑制EMD的端点效应,更好地分解出谐波中含有的不同频率谐波分量。

关键词: 希尔伯特-黄变换, 经验模态分解, 端点效应, 极限学习机, B样条插值, 谐波分析

Abstract: In view of the end effect problems from the experience mode decomposition of Hilbert-Huang Transform process, this paper uses the Extreme Learning Machine(ELM) algorithm to extend the left and the right ends of the original data sequence respectively, takes the mean curve of the data sequence after extension through the B-spline interpolation method. Based on it, the Empirical Mode Decomposition(EMD) process starts, and then abandons both ends of extension data after decomposition is completed, so as to achieve the purpose of restrain end effect. By means of the contrast analysis of indicators of traditional neural network extension and support vector machine extension, the results show that, the algorithm not only can effectively restrain the end effect of EMD process but also get a great advantage in learning speed and decomposition accuracy. Applying the method to harmonic detection in power system, the simulation results show that this method can effectively inhibit the end effects of EMD and make a better decomposition of harmonic components of different frequencies contained in the harmonic signal.

Key words: Hilbert-Huang Transform(HHT), Empirical Mode Decomposition(EMD), end effect, Extreme Learning Machine(ELM), B-spline function, harmonic analysis