基于相似曲线簇和GBRT方法的超短期风电功率预测  

Similar Curve Cluster and GBRT Method-based Ultra-short-term Wind Power Prediction

作　　者：张颖超[1,2] 黄飞 邓华[1,2] 支兴亮 李慧玲 ZHANG Yingchao;HUANG Fei;DENG Hua;ZHI Xingliang;LI Huilin(School of Information and Control,Nanjing University of Information Science&Technology,Nanjing 210044, China;Collaborative Innovation Center on Forecast and Evaluation of Meteorological Disasters,Nanjing University of Information Science&Technology,Nanjing 210044,China)

机构地区：[1]南京信息工程大学信息与控制学院,江苏南京210044 [2]南京信息工程大学气象灾害预报预警与评估协同创新中心,江苏南京210044

出　　处：《华北电力大学学报（自然科学版）》2018年第6期15-20,共6页Journal of North China Electric Power University：Natural Science Edition

基　　金：国家自然科学基金资助项目(41675156);国家公益性行业(气象)科研专项(GYHY20110604);江苏省六大人才高峰项目(WLW-021)

摘　　要：为了减少训练数据的冗余信息,提高风电功率预测的精度,提出了基于相似曲线簇和GBRT方法的超短期风电功率预测模型。首先对历史风速序列进行相似曲线簇的提取,采用相似离度作为相似性判据,对大量历史风速序列与测试集风速序列进行相似性的判断,继而找出相似性好的风速曲线簇以及曲线簇中每个风速点对应的功率,并将其作为最终的训练样本,然后采用梯度提升回归树(GBRT)模型进行风电功率的预测。用上海某风场的数据进行对比试验,结果表明,该方法能够明显提高超短期风电功率预测的精度,具有实际意义。In order to reduce the redundant training data and improve the accuracy of wind power prediction,this paper proposed an ultra-short-term wind power prediction model on the basis of similar curve clusters and GBRT(gradient boosting regression tree)method.Firstly,the similar curve cluster from the historical wind speed sequence was extracted.The similarity degree was taken as the criterion of similarity to determine the similarity between a large number of historical wind speed sequence and test set wind speed sequence.Then the similar curve clusters and the corresponding power of each wind speed point in the curve clusters were picked out to be the final training samples.Finally,the GBRT model was applied to wind power prediction.Compared with the data of a wind farm in Shanghai,this method can improve the accuracy of ultra-short-term wind power prediction,which is of practical significance.

关 键 词：超短期风电功率预测 相似曲线簇 相似离度 GBRT 

分 类 号：TM614[电气工程—电力系统及自动化]