中国电力 ›› 2012, Vol. 45 ›› Issue (4): 78-81.DOI: 10.11930/j.issn.1004-9649.2012.4.78.3

• 新能源 • 上一篇    下一篇

基于小波变换的ARMA-LSSVM短期风速预测

赵辉1, 李斌1, 李彪2, 岳有军1   

  1. 1. 天津理工大学 天津市复杂控制理论与应用重点实验室,天津 300384;
    2. 陕西长岭纺织机电科技有限公司,陕西 宝鸡 721013
  • 收稿日期:2011-08-23 修回日期:2011-12-10 出版日期:2012-04-18 发布日期:2016-02-29
  • 作者简介:赵辉(1963-),男,天津人,博士,教授,从事电力电子控制技术研究。
  • 基金资助:
    国家高技术研究发展计划(863计划)资助项目(2007AA041401); 天津市高等学校科技发展基金资助项目(2006ZD32)

Short-term wind speed forecasting model based on ARMA-LSSVM and wavelet transform

ZHAO Hui1, LI Bin1, LI Biao2, YUE You-jun1   

  1. 1.Tianjin Key Laboratory for Control Theory & Applications in Complicated System, Tianjin University of Technology, Tianjin 300384, China;
    2. Shaanxi Changling Textile Mechanical & Electronic Technology Co., Ltd., Baoji 721013, China
  • Received:2011-08-23 Revised:2011-12-10 Online:2012-04-18 Published:2016-02-29

摘要: 对风电场风速的准确预测,可以有效减轻并网后风电对电网的影响,提高风电市场竞争力。提出将时间序列自回归滑动平均模型(Auto Regressive Moving Average, ARMA) 与最小二乘支持向量机模型(Least Square Support Vector Machine,LS-SVM)相结合的混合模型短期风速预测方法。采用小波变换(Wavelet Transform,WT)方法将历史风速序列分解成具有不同频率特征的序列。根据分解后各分量的特点,对于低频趋势分量选取LS-SVM方法进行预测,而高频波动分量则选取ARMA模型进行预测,采用小波重构得到最终预测结果。仿真实例表明,不同的预测方法整体的预测精度不同,而混合模型预测的均方根误差最低为11.5%,与单一预测方法相比,混合模型提高了预测精度。

关键词: 短期风速预测, 小波变换, 时间序列, 最小二乘支持向量机

Abstract: A wind speed forecasting with high accuracy can effectively reduce or avoid the adverse effect of wind farm on power grids, meanwhile it can enhance the competitive ability of wind power in electricity market. A short-term wind speed forecasting method based on auto-regressive moving average (ARMA) model and least square support vector machine (LS-SWM) model was proposed. By using wavelet transform method, the historical load data was decomposed into series with different frequency characteristics. The low frequency components were predicted by LS-SVM model, while the high frequency components were predicted by ARMA model. The final forecasting results were obtained with wavelet reconstruction. Research results show that the prediction accuracy is different from each method. The mean square error of the proposed hybrid forecast model is 11.5%, better than the results by single forecasting methods.

Key words: short-term wind speed forecasting, wavelet transform, time series, least square support vector machine

中图分类号: