Publication Date


Type of Culminating Activity


Degree Title

Master of Science in Mechanical Engineering


Mechanical and Biomechanical Engineering

Major Advisor

Inanc Senocak, Ph.D.


With installed wind power capacities steadily on the rise, balancing the loads on electrical grids is challenging due to the intermittency of the wind. Short-term wind power forecasting can be a valuable tool for better informing grid operators on the available wind power. Current short-term wind forecasting techniques typically adopt mesoscale weather forecasting models with spatial resolutions on the order of a kilometer. On relatively flat terrain, use of mesoscale models may prove effective, but application to complex terrain induces large forecasting errors. To address this issue, a baseline incompressible flow solver for GPU (graphics processing unit) clusters is extended to simulate neutrally-stable atmospheric flows over complex terrain with the ultimate goal of developing a comprehensive short-term wind fore-casting capability that can resolve winds at turbine hub height. In the extended wind model, the large-eddy simulation (LES) technique with a Lagrangian dynamic subgrid-scale (SGS) model is implemented to better capture the effects of atmospheric turbulence over complex terrain. Additionally, the immersed boundary method (IBM) is adopted to numerically represent the complex terrain on a Cartesian mesh. Validation is performed using common benchmark cases. Performance results obtained from simulating the Bolund Hill Experiment demonstrates that faster than real-time computations are realized with GPU clusters. While the results are encouraging and justifies the foundation for a short-term wind forecasting capability, the work does not account for all factors in wind forecasting and the results can be considered as a first attempt requiring further improvements.