Model-Optimized Dispatch for Closed-Loop Power Control of Waked Wind Farms

At present, wind farms control their power production by using a closed-loop feedback control approach, which distributes the total power to the wind turbines. However, the total power is distributed according to the turbines' available power only. The use of model-predictive control allows considering multiple objectives, nonetheless, since it is open-loop, it can result in poor tracking of the total power reference. This work is the first to combine the standard, closed-loop feedback controller with model-predictive optimization (MPO) in order to yield the benefits of both approaches. As such, we developed an optimization-based dispatch function employed in a closed-loop feedback controller. The dispatch function uses model-predictive, multi-objective optimization to determine the distribution of the total power to the wind turbines. The model employed in the developed dispatch function is the Dynamic Flow Predictor, which uses Kalman-filter-driven feedback to correct the wind farm flow model dynamically. The developed optimization-based dispatch function is compared to dispatch functions commonly employed in present wind farms in a secondary regulation scenario in dynamic simulation. The comparison is carried out on an 80-turbine, large-scale wind farm. The newly developed, optimization-based dispatch function yields a reduction of the mean error and the normalized root-mean-square (NRMS) error by 43% and 36% with respect to the best-performing, commonly used dispatch function. Furthermore, for the large-scale wind farm, the duration of the MPO is only 0.21 s, which is two orders of magnitude faster than comparable approaches in the literature.