Data-driven multi-objective predictive control of offshore wind farm based on evolutionary optimization

Abstract

Abstract A data-driven multi-objective predictive control approach is proposed to increase the power production and reduce fatigue loads on a wind farm level using evolutionary optimization. The FLORIS (FLOw Redirection and Induction in Steady-state) tool is employed to characterize the wake characteristics within a wind farm and generate necessary data for data-driven prediction. A data driven wind farm predictor (WFP) is then constructed by using the turbine yaw angles as inputs and the wind farm power and thrust load as outputs under different inflow wind speeds and wind directions. Based on the WFP, a constrained optimization problem is formulated and the multi-objective predictive controller (MOPC) is designed based on wake steering and evolutionary optimization while considering the yaw angle control constraints. Extensive design experiments are conducted under various wind speeds and wind directions, and the results indicate that the wind farm thrust can be reduced by up to 12.96% while the wind farm power production can be well maintained at almost the same level by using the proposed control in comparison with a conventional model predictive control. The yaw angles optimized from the proposed control are more responsive and active in tuning the wind farm power production and thrust load mitigation than that in the conventional control method.