耦合风速测量的风力机时空尾流重构  

Spatiotemporal wake field reconstruction of wind turbine coupled with wind speed measurements

作　　者：王龙滟[1] 陈梦[1,2] 袁建平 WANG Longyan;CHEN Meng;YUAN Jianping(National Research Center of Pumps,Jiangsu University,Zhenjiang,Jiangsu 212013,China;Changzhou Borui Electric Power Automation Equipment Co.,Ltd.,Changzhou,Jiangsu 213025,China)

机构地区：[1]江苏大学国家水泵及系统工程技术研究中心,江苏镇江212013 [2]常州博瑞电力自动化设备有限公司,江苏常州213025

出　　处：《排灌机械工程学报》2025年第3期260-267,共8页Journal of Drainage and Irrigation Machinery Engineering

基　　金：国家自然科学基金资助项目(12002137);江苏省博士后基金资助项目(2021K110B);江苏大学高级人才启动基金资助项目(20JDG065)。

摘　　要：为了测量风力机动态尾流详细流场信息,提出了一种物理信息深度学习方法来解决这一困境,该方法通过结合少量稀疏的激光雷达测量和流体动力学方程(Navier-Stokes方程组)来实现动态尾流的时空重构.具体而言,神经网络中嵌入流体动力学方程组来约束输出的物理量,包括顺风速度、侧风速度和压强,保证输出的可解释性和合理性.以偏航过程中的动态尾流为案例,仅提供少量尾流内部测量数据作为训练集,完成了局部动态尾流的时空重构.该方法成功捕捉偏航过程中尾流动态变化的趋势,准确地预测尾流轨迹和偏转量.在全局尾流的时空重构中,所提出方法完整还原真实流场的流动演变过程,这项技术在风电场智能控制方面具有巨大的潜力.To measure the detailed flow field information of the dynamic wake of a wind turbine,a deep learning method for physical information was proposed to solve this dilemma,which combined a small number of sparse measurements and fluid dynamics equations to achieve spatiotemporal reconstruction of dynamic wake.Specifically,Navier-Stokes equations were embedded in the neural network to constrain the output physical quantities,including downwind speed,crosswind speed and pressure,to ensure the explainability and rationality of the output.Taking the dynamic wake during yaw as an example,only a small amount of internal wake measurement data is used as a training set,and the spatiotemporal reconstruction of the local dynamic wake is completed.This method successfully captures the dynamic trend of wake flow during yaw,and accurately predicts the wake trajectory and deflection.In the spatiotemporal reconstruction of global wake,the proposed method completely restores the flow evolution process of the real flow field,which has great potential in the intelligent control of wind farms.

关 键 词：水平轴风力机 动态尾流 离散测量 深度学习 时空重构 

分 类 号：S277.9[农业科学—农业水土工程]