Towards First-Principles Based Wake and Wake Interaction Models for Wind-Farm Layout Optimization

主要研究者：Sreenivas，Kidambi提案编号：1236124机构：田纳西大学查塔努加标题：走向第一原理为基础的尾流和尾流相互作用模型的风电场布局优化风能已成为一个有前途的可再生能源多年来，大型风电场正在建设世界各地。这些风电场利用多兆瓦风力涡轮机产生数百兆瓦（MW）的电力。风力涡轮机的操作受许多因素影响，然而，从其它涡轮机吸入尾流导致功率输出降低和涡轮机叶片上的动态负载增加。为了最小化风电场的成本，涡轮机通常位于非常接近的位置，导致风电场中许多尾流的相互作用和合并。以往的研究表明，由于尾流的估计功率损失范围从5%到20%，这取决于风力发电场的布局。虽然风力涡轮机尾流的建模一直是近30年的研究课题，尾流相互作用的分析通常使用简单的尾流模型，预测速度和尾流中的其他流动变量，使用一组显式方程或解析表达式。今天？的计算资源，现在可以通过求解Navier-Stokes方程，而无需引入尾流模型，对具有旋转叶片的多个风力涡轮机进行高分辨率模拟。虽然这种方法可以提供具有尾流相互作用的风力涡轮机流场的更高保真度的表示，但计算机运行时间将从几天到几周不等。因此，需要具有改进的场模拟模型和尾流模型，其既高保真又足够经济以执行有效设计和优化风电场布局所需的大量模拟情况，以便最小化尾流摄入的不利影响。此外，这种组合模型可以用于通过与不断更新的风气候（速度和方向）预测相结合来改进对现有风电场的电力生产的估计。准确的发电量预测对于风电场的经济运行和减少电网稳定性问题是必要的。本项目将探索两种新的、高保真的、低成本的风力涡轮机尾流和尾流相互作用计算模型。首先是一个新的抛物型Navier-Stokes模型的基础上的一次/二次流近似，允许空间行进的解决方案，而无需任何近似的压力梯度。它可以用于定常模式与时间无关的涡轮盘模型或非定常模式与涡轮叶片模型。它将改进现有的模型，不仅适用于远尾流，而且适用于近尾流区域。它的计算成本模拟通过风力发电场的流量将大大低于（数量级）比纳维尔-斯托克斯模拟。 尾流和尾流相互作用的第二个模型将涉及使用诸如本征正交分解或动态模式分解之类的技术将Navier-Stokes模拟分解成模式。与抛物型Navier-Stokes模型相比，该模型将显著降低计算成本。在风电场布局优化领域，本研究将提出将全局优化算法与基于梯度的算法相结合的方法，以减轻其不足并减少优化的总体挂钟时间。SimCenter是田纳西大学查塔努加分校的综合研究和教育中心，参与了广泛的STEM活动，以支持激发K-12学生对科学和数学相关学科的兴趣。它经常在一年中接待数十组K-12学生。它还维护一个专门提供STEM相关资源的网站。作为拟议工作的一部分，这些活动将扩大，以满足当地博物馆对合作的兴趣，以接触更广泛的观众。一个风力发电场的桌面模型将被创建和使用的一部分，图尔斯进行。这将使学生有实际操作经验与功能风力涡轮机。高中生、本科生和研究生将参与这项活动的各个方面。高中生和本科生将参与开发桌面模型，并将为SimCenter网站开发软件应用程序。此外，本科生将协助生成模拟所需的计算网格，研究生将实施本提案中概述的方法。

PI: Sreenivas, KidambiProposal Number: 1236124Institution: University of Tennessee ChattanoogaTitle: Towards First-Principles Based Wake and Wake Interaction Models for Wind-Farm Layout OptimizationWind energy has become a promising renewable energy source over the years with large wind farms being built worldwide. These wind farms produce hundreds of megawatts (MW) of power utilizing multi-megawatt wind turbines. The operation of a wind turbine is affected by many factors, however the ingestion of wakes from other turbines results in reduced power output and increased dynamic loading on the turbine blades. In an effort to minimize the cost of the wind farm, turbines are typically located in close proximity, resulting in the interaction and merging of many wakes in the wind farm. Previous studies indicate that estimated power losses due to wakes can range from 5% to 20%, depending upon the wind farm layout.Although modeling of wind turbine wakes has been a research topic for nearly 30 years, analysis of wake interactions has typically used simple wake models that predict velocity and other flow variables in the wake using a set of explicit equations or analytical expressions. With today?s computational resources it is now possible to perform high-resolution simulations for multiple wind turbines with rotating blades by solving the Navier-Stokes equations without introducing a wake model. Although this approach would provide a higher fidelity representation of wind-turbine flow with wake interaction, the computer run time would range from days to weeks. Thus, there is need to have improved field simulation models and wake models that are both high fidelity and sufficiently economical to perform the large number of simulation cases required to effectively design and optimize the layouts of wind farms so as to minimize the adverse affects of wake ingestion. Additionally, such a combined model could be used to improve estimates of power production from existing wind farms by integration with continually updated wind climate (speed and direction) forecasts. Accurate power production forecasts are necessary for economical operation of the wind farm and to reduce power-grid stability issues.This project will explore two new, high-fidelity, low-cost computational models for wind-turbine wakes and wake interactions. The first is a new Parabolic Navier-Stokes model based on a primary/secondary flow approximation that allows spatial marching solution without any approximation for pressure gradients. It can be used in steady mode with time-independent turbine-disk models or in unsteady mode with rotating-blade models. It will improve upon existing models by being applicable not only in the far wake but also in the near-wake region. Its computational cost for simulating flow through a wind farm would be substantially lower (orders of magnitude) than a Navier-Stokes simulation. The second model for wake and wake-interaction will involve decomposing Navier-Stokes simulations into modes using techniques such as Proper Orthogonal Decomposition or Dynamic Mode Decomposition. This model will significantly reduce the computational cost compared to the Parabolic Navier-Stokes model. In the area of wind farm layout optimization, this study will address methods for coupling a global optimization algorithm with a gradient-based algorithm to mitigate their deficiencies and reduce overall wall-clock time for optimization.The SimCenter, an integrated research and education center within the University of Tennessee at Chattanooga, has been involved in a broad range of STEM activities that support the goal of stimulating interest among students in K-12 in science and math related disciplines. It regularly hosts dozens of groups of K-12 students over the course of the year. It also maintains a website dedicated to STEM related resources. As part of this proposed effort, these activities will be expanded to respond to interest from local museums in collaborations to reach a wider audience. A table-top model of a wind farm will be created and used as part of the tours that are conducted. This will enable students to have hands-on experience with functioning wind turbines. High school, undergraduate and graduate students will be involved in various aspects of this activity. The high-school and undergraduate students will be involved in developing the table-top model and will also develop software applications for the SimCenter website. In addition, undergraduate students will assist in the generation of computational grids required for the simulations, and graduate students will implement the approaches outlined in this proposal.