基于变量选择深度信念神经网络的风速预测  

Wind Speed Prediction Based on DBN with Variable Selection

作　　者：李大中[1] 李昉 张克延 LI Dazhong;LI Fang;ZHANG Keyan(School of Control and Computer Engineering,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学控制与计算机工程学院,河北保定071003

出　　处：《华北电力大学学报（自然科学版）》2021年第1期62-68,共7页Journal of North China Electric Power University：Natural Science Edition

基　　金：河北省自然科学基金资助项目(F2011502001).

摘　　要：风速波动具有随机性和不确定性,导致风速预测的准确度不高。准确的风速预测对于优化风电运行策略和提高发电效率具有重大意义。利用最大信息系数(Maximal Information Coefficient,MIC)对风机SCADA数据进行变量相关性分析,并以MIC值大小对原始变量排序,将包含7项变量的子集作为深度信念网络(Deep Belief Network,DBN)输入,得到MIC-DBN风速预测模型。基于风场实际数据将MIC-DBN模型与BP神经网络模型和GA-BP模型进行测试对比,实验结果表明,MIC-DBN风速预测模型的预测精度和泛化性能具有良好的效果。Accurate wind speed prediction is of great significance for optimizing wind power operation strategies and improving power generation efficiency.However,wind speed fluctuation is random and uncertain,making wind speed prediction inaccurate.This paper utilized the maximum information coefficient(MIC)to analyze the variables correlation of the SCADA data of the wind turbine,sorted the original variables according to the MIC value.We took the subset with 7 variables as the deep belief network(DBN)input to get the MIC-DBN wind speed prediction model.Based on the actual wind filed data,we compared the MIC-DBN model with the BP neural network model and the GA-BP model.The experimental results show that the MIC-DBN model performs well in accurate prediction and generalization.

关 键 词：风场大数据 变量选择 深度信念网络 受限玻尔兹曼机 最大信息系数 

分 类 号：TM614[电气工程—电力系统及自动化]