Efficient Simulation of Wind and Wake Flows in Wind Farms Using a Preconditioned Multistage Solver

Abstract

The simultaneous simulation of microscale wind and wind turbine wake is attempted in this work using a preconditioned multistage solver and an immersed wind turbine model. This work demonstrates a low Mach number preconditioning formulation for high Reynolds number, low Mach number atmospheric flows. The preconditioning is implemented together with multigrid approach into a multistage solver in order to provide an efficient scheme, which allows for the routine use of computational fluid dynamics in simulations of the atmospheric flow and wakes within wind farms. Microscale wind simulations are performed using the preconditioned solver over several test cases including Askervein Hill, Kettles Hill and Bolund Hill to demonstrate the superior convergence, accuracy and robustness of the method. In addition, the performance of RANS solver is assessed in prediction of microscale wind variations over topography with varying complexity. Next, A novel immersed wind turbine model is developed and used to simulate the evolution of single and multiple wake interactions. The model is formulated in order to reduce the stringent grid requirements for resolving the blade boundary layer and near-wake region behind the wind turbines. The model is first evaluated by comparing the predicted evolution of the velocity and turbulence intensity in the far wake with measurements performed in a wind tunnel. Predictions are also compared with full-scale measurements of a single wake at the Sexbierum wind farm. The performance of the model is also assessed in predicting wake interaction in the same wind farm by comparing the power performance with measurements in an operating wind farm. The agreement between the model results and measurement for all cases is satisfactory for both single and double wake predictions. In next step, the simulations are performed over offshore wind farm Lillgrund and sensitivity of power loss to wind direction is investigated in two first rows of the farm. The results of simultaneous wind and wake flow are also presented over two wind farms located in complex terrain in Spain and Switzerland. The recovery rate of wake and turbulence characteristics in wind farm caused by the topography are assessed. Overall the results of array loss in wind farm located in complex terrain demonstrate the sensitivity of wind farm performance in complex terrain to small variations in wind direction. The model yield acceptable results for all test cases, however, to justify the additional computational cost of RANS simulations compared to simplified engineering models, the accuracy needs to be further improved and the computational cost must be reduced. Overall, the results over various test cases demonstrates the capability of the model to resolve the wake interaction in wind farms and to estimate the array loss for different arrangements of turbines and incoming wind directions. In addition to reducing the computational