CAREER: Multi-scale modeling of short-term forecasting and grid integration of wind energy over complex terrain

项目负责人：Inanc senocak提案编号：1056110机构：博伊西州立大学职称：职业：复杂地形上风能短期预测和电网整合的多尺度建模增加风能资源用于发电的兴趣日益增长。但是，提高风能在整体能源生产中的比例要比简单地在多风的平坦地区安装风力发电场复杂得多。本研究的总体目标是利用多尺度建模方法，更好地了解不同大气稳定性条件下复杂地形峡谷湍流的特征，从而可靠地利用这些风能资源用于发电。提出了一种将大气过程从中尺度降至可解析复杂地形特征的微观尺度的智能MeritA多尺度模拟方法。具体来说，提出了一种计算速度快的风模拟能力，利用图形处理单元（GPU）集群上的加速计算流体动力学（CFD）模拟，预测涡轮机水平短期（例如0-6小时）的风。将大涡模拟技术与动力学过程相结合，应用于复杂地形大气流动的高分辨率研究。提出了一种支持新兴GPU集群多尺度耦合计算的仿真引擎。在这个引擎中，一个中尺度大气模式（例如天气研究与预报模式）将在中央处理器上执行。微尺度地形解析CFD模型将与中尺度模型同时在GPU上执行，以提供涡轮水平的高分辨率模拟。微尺度模拟还将用于改进中尺度模式中的表层参数化，以在模拟模式之间建立一个半耦合。中尺度和微观尺度模式将同时执行，以预测风能生产的复杂地形上的风。模拟结果将用于湍流动能收支分析，这将有助于识别和了解在不同稳定性条件下复杂地形环境中湍流产生的主要来源。风速的标准偏差将被检查并与现有的相关性进行比较。模拟将根据从爱达荷州的一个测试区域获得的测量结果进行验证，该测试区域地形复杂，配有气象站仪器。将研究实验和计算数据，以找到合适的缩放参数来表征峡谷内和山脊内的流动结构。更广泛的影响在短期风电预测和电网整合方面存在巨大的技术差距。提高各类地形短期预报的准确性对风能产业具有重要意义。预计在项目结束时，将更好地了解复杂的地形流动，用于在这些风环境中进行短期预测。拟议的教育和推广活动集中在超级计算和应用于风能的科学可视化上，作为激发学生对计算科学兴趣的一种手段。一个配备了平铺显示可视化集群的计算建模实验室将用于研究和教育。本实验室将建立一个gpu加速的教育CFD模型，并提供一套教程，以支持热学和流体科学教育，用于流体力学课程。加速计算平台将使本科生和研究生能够快速完成计算密集型模拟，以研究底层物理。gpu加速教育CFD模型的免费版本也将向公众开放。将开发一个超级计算机展台，用于博伊西州立大学正在进行的K-12外展项目（例如e-Girls， e-Camp, e-Day）。模拟展台将包括在复杂地形中建模和模拟风的实践练习，以及来自地球和空间科学的科学数据和高分辨率图像的可视化。

PI: Inanc SenocakProposal Number: 1056110Institution: Boise State UniversityTitle: CAREER: Multi-scale modeling of short-term forecasting and grid integration of wind energy over complex terrainThere is a growing interest to increase the utilization of wind energy resources for electricity production. But increasing the percentage of wind energy in overall energy production is much more complex than simply installing wind farms in windy areas with flat terrains. The overall goal of this research is to better understand the characteristics of turbulent flows in complex terrain with canyons under different atmospheric stability conditions using a multi-scale modeling approach, so that these wind energy resources can be reliably harnessed for the production of electricity. Intellectual MeritA multi-scale modeling approach that connects atmospheric processes at the meso-scale down to the micro scale where complex terrain features can be resolved is proposed. Specifically, a computationally fast wind simulation capability is proposed to forecast winds at the turbine level for the short-term (e.g. 0-6 hours), using accelerated computational fluid dynamics (CFD) simulations on graphics processing units (GPU) clusters. The large eddy simulation (LES) technique with the dynamic procedure will be applied to study atmospheric flows in complex terrain with high resolution. A simulation engine is proposed to support multi-scale coupled computations on emerging GPU clusters. In this engine, a meso-scale atmospheric model (e.g. the Weather Research and Forecasting model) will be executed on central-processing units (CPU). A microscale terrain-resolving CFD model will be executed on the GPU concurrently with the meso-scale model to provide high resolution simulations at the turbine level. The micro-scale simulations will also be used to improve the surface layer parameterizations in the meso-scale model to establish a one-and -a-half-way coupling between the simulation models. The meso and micro scale models will be executed concurrently to forecast winds over complex terrain for wind energy production. Simulation results will be used to perform turbulent kinetic energy budget analysis, which will help identify and understand the primary source of turbulence production in complex terrain environment under different stability conditions. Standard deviations of the wind velocity will be examined and compared against existing correlations. Simulations will be validated against measurements obtained from a test area in the state of Idaho in a complex terrain area instrumented with weather stations. Both experimental and computational data will be studied to find appropriate scaling parameters to characterize the flow structure within canyons and along ridges. Broader ImpactsSubstantial technology gaps exist in short-term wind power forecasting and grid integration. Improving the accuracy of short-term forecasts for all types of terrain is of great importance to wind energy industry. A better understanding of complex terrain flows used to develop short-term forecasts in these wind environments is expected at the end of the project. The proposed education and outreach activities are centered on supercomputing and scientific visualization applied to wind energy as a means to motivate student interest in the computational sciences. A computational modeling laboratory equipped with a tiled-display visualization cluster will be used for research and education. A GPU-accelerated educational CFD model with a tutorial set to support thermal and fluid sciences education will be set up in this laboratory for use in fluid mechanics courses. The accelerated computational platform will enable both undergraduate and graduate students to finish compute-intensive simulations quickly to investigate the underlying physics. A free version of GPU-accelerated educational CFD model will also be made publically available. A supercomputing booth will be developed for use at on-going K-12 outreach programs at Boise State University (e.g. e-Girls, e-Camp, e-Day). The simulation booth will include hands-on exercises for modeling and simulation of winds in complex terrain and visualization of scientific data and high resolution imagery from earth and space sciences.