Wind turbine modeling for computational fluid dynamics

Abstract

With the shortage of fossil fuel and the increasing environmental awareness, wind energy is becoming more and more important. As the market for wind energy grows, wind turbines and wind farms are becoming larger. Current utility-scale turbines extend a signi cant distance into the atmospheric boundary layer. Therefore, the interaction between the atmospheric boundary layer and the turbines and their wakes needs to be better understood. The turbulent wakes of upstream turbines affect the flow field of the turbines behind them, decreasing power production and increasing mechanical loading. With a better understanding of this type of flow, wind farm developers could plan better-performing, less maintenance-intensive wind farms. Simulating this flow using computational fluid dynamics is one important way to gain a better understanding of wind farm flows. In this study, we compare the performance of actuator disc and actuator line models in producing wind turbine wakes and the wake-turbine interaction between multiple turbines. We also examine parameters that a ect the performance of these models, such as grid resolution, the use of a tip-loss correction, and the way in which the turbine force is projected onto the flow field. We see that as the grid is coarsened, the predicted power decreases. As the width of the Gaussian body force projection function is increased, the predicted power is increased. The actuator disk and actuator line models produce similar wake profiles and predict power within 1% of one another when subject to uniform in flow. The actuator line model is able to capture flow structures near the blades such as root and tip vortices, which the actuator disk does not capture, but in the far wake, they look very similar. The actuator line model was validated using the wind tunnel experiment conducted in The Norwegian University of Science and Technology, Trondheim. Good agreement between the model and the experiments was obtained, with maximum percentage difference in power coefficients of 25% and 40% for thrust coefficient. The actuator line and actuator disk models were compared when running large-scale wind farm simulations. Normalized power was very similar for both models but dimensional power was within 1 and 17% difference from of each other. The actuator disk model was able to run roughly 3 times faster though. This work shows that actuator models for wind turbine aerodynamics are a viable alternative to fully blade resolving simulations. However, care must be taken to use the proper grid resolution and force projection to the CFD grid to obtain accurate predictions of aerodynamic forces and hence power. More work needs to be done to determine the best method of body force projection onto the CFD grid.