基于膨胀腐蚀聚类方法的风电功率预测  

WIND POWER PREDICTION BASED ON DILATION AND EROSION CLUSTERING METHOD

作　　者：周晓 冬雷[1] 郝颖[1] 廖晓钟[1] 高阳 Zhou Xiao;Dong Lei;Hao Ying;Liao Xiaozhong;Gao Yang(College of Automation,Beijing Institute of Technology,Beijing 100081,China;College of Electric Power,Shenyang Institute of Engineering,Shenyang 110136,China)

机构地区：[1]北京理工大学自动化学院,北京100081 [2]沈阳工程学院电力学院,沈阳110136

出　　处：《太阳能学报》2018年第12期3536-3543,共8页Acta Energiae Solaris Sinica

基　　金：国家自然科学基金(51607009)

摘　　要：提出一种基于膨胀腐蚀的聚类方法,并利用UCI(university of california irvine)数据集进行实验仿真证明此方法的可行性。将此聚类方法应用于风电功率预测中的NWP(numerical weather prediction)信息分类,选择与预测日同一类的历史日数据作为训练样本,利用广义回归神经网络预测功率,并与直接预测的方法相比较,仿真结果表示基于膨胀腐蚀对历史日数据分类后再预测的精度较高。A new clustering method based on dilation and erosion is proposed and UCI (University of California Irvine) data set is used to carry out experimental simulation to prove the feasibility of this method.Then this clustering method is used to classify NWP (numerical weather prediction)information in wind power prediction,selecting the historical day data of the same type as the forecast day data as the training sample,and the generalized regression neural network is used to predict the power and compare with the direct prediction method.The simulation results show that the re-prediction has higher prediction accuracy after classification of historical day data based on dilation and erosion clustering analysis.

关 键 词：风电功率预测 聚类 膨胀腐蚀 广义回归神经网络 

分 类 号：TK614[动力工程及工程热物理—生物能]