Closed-loop Wind farm control for floating turbines via online yaw and induction setpoint optimization

Abstract This paper investigates closed-loop wind-farm control for floating wind turbines via online setpoint optimization of yaw setpoints for wake steering and turbine derating for induction control. A FLORIS-based supervisory controller is coupled to a mid-fidelity floating wind-farm plant simulated in FAST.Farm through a ZeroMQ interface. Model–plant mismatch is handled through two alternative, separately evaluated approaches: (i) calibration of the FLORIS wake and turbulence response against FAST.Farm (including a rank-dependent attenuation to match downstream wake-added turbulence trends), and (ii) a turbine-wise Gaussian Process Regression (GPR) residual model enabled in selected closed-loop (CL) cases. A four-turbine weakly staggered farm is assessed under turbulent inflow and operational events (turbine shutdowns and time-varying wind direction), and closed-loop performance is benchmarked against lookup-table policies. Results show consistent power gains over greedy control (approximately 10—12% for wake steering and about 16% for induction in the tested condition), with CL providing systematic advantages under shutdown scenarios, while GPR yields mixed benefits under turbulent CL operation.