Page 126 - 201806

第 37 卷 第 6 期 Vol. 37, No.6
2018 年 11 月 Journal of Applied Acoustics November, 2018

⋄ 研究报告 ⋄

基于数据融合的风电机组噪声预测

张智勇 1 余 金 2† 常 鹏 2 徐其丹 2 李 阳 2

(1 喀什大学物理与电气工程学院 喀什 844008)

(2 国网新疆电力有限公司经济技术研究院 乌鲁木齐 830049)
摘要 根据风电机组噪声信号检测复杂的情况，研究风电机组非声学参数的信息熵特征，对机组噪声进行多
源数据融合预测。分析基于信息熵的非声学参数的特征提取方法，并对传统的基于遗传算法的支持向量机回
归的缺陷提出改进，结合实际应用的非声学参数的信息熵特点平衡遗传算法 (GA) 的终止条件。通过统计分
析完成了输入变量的筛选，去除了对预测影响较大的共线性因素，并实现了输入降维提高预测精度和速率。最
后应用数据的信息熵特征，训练改进的遗传算法的支持向量机回归，建立最终的多源数据特征级融合预测模
型。通过对比表明基于多源数据融合的预测方法精度最高，预测结果的相对误差平均值为 0.7757%，具有实际
可行性。
关键词 风力发电，噪声预测，多源数据融合，支持向量机回归，遗传算法
中图法分类号: TM614 文献标识码: A 文章编号: 1000-310X(2018)06-0956-07
DOI: 10.11684/j.issn.1000-310X.2018.06.018

Wind turbines noise prediction based on data fusion

ZHANG Zhiyong 1 YU Jin 2 CHANG Peng 2 XU Qidan 2 LI Yang 2

(1 School of Electrical Engineering, Kashgar University, Kashi 844008, China)
(2 State Grid Economic and Technical Research Institute Co., LTD. of Xinjiang, Urumqi 830049, China)

Abstract According to the complex situation of wind turbine noise signal detection, the information entropy
characteristics of non-acoustic parameters of wind turbine are analyzed and the noise is predicted by multi-
source data fusion. The feature extraction method based on entropy of non-acoustic parameters is analyzed,
and the traditional method based on the support vector regression of genetic algorithm (GA-SVR) is improved.
While the termination condition of the genetic algorithm is balanced by the information entropy characteristics
of the non-acoustic parameters. What’s more, the input variables are filtered out through the statistical
analysis, removing the collinear factors that have a greater impact on the forecast and realizing the variable
dimension reduction to improve the prediction accuracy and speed. Finally, the information entropy features of
the data are applied to train the improved GA-SVR to establish the final prediction model of multi-source data
feature level fusion. The comparison shows that the accuracy of the prediction method based on multi-source
data fusion is the highest, and the relative error of the predicted results is 0.7757%, which is feasible.
Key words Wind power, Noise prediction, Multi-source data fusion, Support vector regression, Genetic
algorithm

2018-01-27 收稿; 2018-05-01 定稿
作者简介: 张智勇 (1989- ), 男, 内蒙古人, 硕士研究生, 研究方向: 可再生能源及其控制技术。
† 通讯作者 E-mail: 21858309@qq.com