Data driven learning model predictive control of offshore wind farms

This paper presents a data-driven control approach for maximizing the total power generation of the offshore wind farm by using arecently developed learning model predictive control (LMPC) algorithm. The control is designed by coordinating yaw angle control actions of wind turbines to mitigate the wake interactions among the turbines for increasing the total farm power production, which is termed as wake redirection. This paper mainly focuses on designing the architecture and methodology of the LMPC for wind farm, including a unified wind turbine wake interaction model, the LMPC for minimizing an iteration cost function, therecursive feasibility, stability and convergence analysis. Extensive comparative studies are conducted to verify the performance of the LMPC in comparison with the existing model predictive control (MPC) method under the same wind speed conditions. The results show that the wind farm yields up to 15% more power production by using the LMPC than the conventional MPC.

Automated summary

This paper introduces a data-driven control approach for maximizing the total power generation of offshore wind farm using the recently developed learning model predictive control (LMPC) algorithm. The strategy focuses on reducing wake interactions among wind turbines to increase power production, showing a 15% improvement over traditional MPC methods.