基于OS-ELM和Bootstrap方法的超短期风电功率预测  

Ultra-short-term Wind Power Prediction Based on OS-ELM and Bootstrap Method

作　　者：王焱[1] 汪震[1] 黄民翔[1] 蔡祯祺[2] 杨濛濛[1] 

机构地区：[1]浙江大学电气工程学院,浙江省杭州市310027 [2]国网上海市电力公司,上海市200122

出　　处：《电力系统自动化》2014年第6期14-19,122,共7页Automation of Electric Power Systems

基　　金：国家高技术研究发展计划(863计划)资助项目(2011AA050204);国家自然科学基金资助项目(51277160)~~

摘　　要：提出了一种基于在线序贯极限学习机(OS-ELM)的超短期风电功率预测方法。利用OSELM学习速度快、泛化能力强的优点,将批处理和逐次迭代相结合,不断更新训练数据和网络结构,实现了对数值天气预报风速的快速实时修正和风电机组输出功率的快速预测。随后,采用计算机自助(Bootstrap)法构造伪样本,给出了预测功率的置信区间评估。实例和研究结果表明,该预测方法与反向传播(BP)网络、支持向量机(SVM)方法相比,在计算时间上更能满足在线应用需求,而且预测精度相当,有较好的应用前景。An ultra-short-term wind power prediction method based on an online sequential extreme learning machine （OS- ELM） is proposed. Firstly, the OS-ELM is utilized to correct the predicted wind speed sequence so as to amend and improve the accuracy of predicted wind speed. Then, by combining batch processing with successive iteration, real-time prediction of wind turbine power output is accomplished with the help of the advantages of OS-ELM＇ s fast learning speed and strong generalization ability. Finally, a Bootstrap method is adopted to estimate the predicted intervals by resampling data. Analysis results show that, compared with the back propagation （BP） network and support vector machine （SVM） method, this prediction method can better meet the demand of online application and has good application prospects, while its forecasting accuracy is comparable to BP network and SVM method.

关 键 词：风电预测 风速修正 误差区间估计 极限学习机 BOOTSTRAP方法 

分 类 号：TM614[电气工程—电力系统及自动化]