耦合垂直风廓线的机器学习风速订正模型  

Wind speed correction model based on machine learning coupled with vertical distribution of wind speed

作　　者：张流杰 王强[1] 明轩萱 杨树林 叶时彤 王凯[1] 罗坤[1] 樊建人[1] ZHANG Liujie;WANG Qiang;MING Xuanxuan;YANG Shulin;YE Shitong;WANG Kai;LUO Kun;FAN Jianren(State Key Laboratory of Clean Energy Utilization,Zhejiang University,Hangzhou 310027,China)

机构地区：[1]浙江大学能源高效清洁利用全国重点实验室,浙江杭州310027

出　　处：《能源工程》2025年第1期48-54,共7页Energy Engineering

基　　金：国家自然科学基金项目(52206281);浙江省自然科学基金资助项目(LY24E060002)。

摘　　要：研究提出了一种在风速垂直分布规律约束下训练得到机器学习风速订正的方法(耦合模型)以提高模型泛化性能。试验中的机器学习模型采用CNN_LSTM模型,风速垂直分布规律指一段时间各高度的时均风速满足指数分布规律,通过物理与数据结合的方法提升订正效果。研究表明,以测风数据为基准,耦合模型订正结果更接近测风数据,较WRF模拟风速在试验中均方根误差最高可降低1.74m/s,较CNN_LSTM订正模型均方根误差最高可再降低0.46m/s,两种不同的订正方法可将相关系数由0.65左右提升到0.9左右,其中耦合模型相关系数能够提升到0.92左右。本研究提出的通过在机器学习模型训练中耦合风速垂直分布规律的方法,能够有效提高订正模型泛化能力。In this study,a method is proposed for training a machine learning model for wind speed correction under the constraint of vertical wind speed distribution patterns(coupled model)to improve model generalization performance.The CNN_LSTM model is adopted in the experiment for the machine learning model.The vertical wind speed distribution pattern refers to the exponential distribution pattern where certain rules are satisfied by average wind speeds at different heights over a period of time.This method combines physics and data to enhance correction effectiveness.It is indicated by the research that,compared to using measured wind data as a baseline,the results from coupled model are closer to the measured data.The root-mean-square error of the coupled model is up to 1.74 m/s lower than that of WRF simulated wind speed and up to 0.46 m/s lower than that of CNN_LSTM.Two different correction methods can increase the correlation coefficient from 0.65 to 0.9,and the correlation coefficient of the coupled model can be increased to about 0.92.The method proposed in this study,where the vertical wind speed distribution pattern is coupled with machine learning model training,effectively improves the generalization capability of the correction model.

关 键 词：风速订正 机器学习 垂直风廓线 耦合模型 

分 类 号：TK83[动力工程及工程热物理—流体机械及工程]