近地层风的多尺度精细化模拟方法研究

The refined prediction of the low-level wind is closely related to our daily lives, however, precisely simulating the low-level wind becomes very difficult due to the surface friction and the inherent properties of turbulence process. In this study, the MCV (Multi-moment Constraint Finite Volume) method which is more easily to construct high order interpolation over single grid with relatively lower computational burden, is intended to develop a novel CFD (Computational Fluid Dynamics) model by solving Reynolds-averaged Navier–Stokes and the related turbulence equations. Then this MCV based CFD model will be coupled with the meso-scale WRF (Weather Research and Forecasting) model to build a multi-scale modelling system to simulate the low-level wind with high resolution and accuracy at target regions. The needed information for running the CFD model is derived from the meso-scale WRF model. On one hand, using the results of this multi-scale modelling system could study how to simulate the wind over very complex terrain or other underlying surface as well as improve the multi-scale model itself. On the other hand, the success of this study will contribute to the development of the current numerical weather prediction and provide scientific instructions to real wind farms to improve the wind energy forecasting as well as the management of urban air pollution.

近地层风的精细化模拟与人类生活息息相关，但由于地表摩擦影响和近地层大气的湍流运动性质，近地层风成为精细数值模拟的难点。本研究拟采用多矩约束（MCV）方法构造高阶精度且计算经济的局地高分辨率雷诺平均Navier-Stokes计算流体动力模式，并与高分辨率中尺度WRF模式耦合，由中尺度模式提供环境动力场，驱动计算流体(CFD)模式获得高精度的近地层风场模拟，实现对复杂地形区域近地层风的多尺度、高精度模拟，探讨复杂地形和下垫面上风场精细化模拟方法，为实际风电场风能预测以及城市污染物应急管理提供科学依据。

近地层风的精细化模拟与人类生活息息相关，但由于地表摩擦影响和近地层大气的湍流运动性质，近地层风成为精细数值模拟的难点。本研究首先将中尺度数值模式WRF与基于传统有限体积方法的CFD模式OpenFOAM进行耦合并对结果进行验证。之后利用多项式混沌展开(PCE)方法对耦合模式中的不确定性定量化以及机器学习方法改进模式输出风场。最后采用多矩约束(MCV)方法构造了高阶精度且计算经济的局地高分辨率雷诺平均Navier-Stokes计算流体动力模式，并与高分辨率中尺度WRF模式耦合。主要结果表明：.（1）多尺度耦合模式可克服单一中尺度数值模式无法准确地得到近地面尤其是复杂下垫面情形下的风场问题。通过对研究区域2019年7月的风场的模拟及检验，发现耦合模式对于偏差(ME)和均方根误差(RMSE)的改善分别为93.08%和36.43%。.（2）不确定性定量化分析表明模式初始场的不确定性对风场预测精度的影响要大于内在湍流模型。基于模式多种气象变量（风速、风向、气压、气温和湿度）和风场观测构建的机器学习模型可在很大程度上改善数值模式短期（24 h）预测。.（3）将在局地高阶重构、数值耗散和稳定性方面具有优势的MCV构建的CFD模型和中尺度WRF结合后的数值结果优于基于传统有限体积方法的WRF-OpenFOAM。.该项目成果对复杂下垫面下近地层风的高精度模拟和预测有重要的指导意义，可为实际风电场风能预测以及城市污染物应急管理提供科学依据。