Stochastic wind farm flow generation using a reduced order model of LES

Abstract

Large eddy simulations are computationally expensive, which typically limits the simulation time and hence the statistics useful for validation. Here, a reduced order model with a stochastic engine is used to generate stochastic inflow to 48 wind turbines in the Lillgrund wind farm. The stochastic flow realization captures the power production and more importantly the correct cross-correlations in power and blade root flapwise bending moments over the large spatiotemporal region, which the wind farm covers. The analysis show how the cross-correlation increase further into the wind farm, where the highly turbulent flow becomes increasingly self-organized and how the dynamic interaction across the entire wind farm is captured by the first eight modes. Overall, the reduced order model captures the correct dynamics at a fraction of the computational cost, effectively generalizing the LES results beyond a relatively short simulation period.