In the control system of wind turbines, model predictive control (MPC), as an advanced control method, has gained increasing attention. However, modeling the wind turbine in MPC is a major challenge. This paper presents a linear time-varying (LTV) modeling and validation framework for wind turbines toward MPC. A state-space model is constructed from turbine drivetrain and aerodynamic mechanisms, where the rotor speed, generator torque, and pitch angle are selected as the core states. The aerodynamic sensitivities are updated online using coefficient maps, enabling the model to reflect operating-point-dependent dynamics. To assess model fidelity, key parameters are identified and compared using Bladed simulation data under both free-running and controlled operating conditions. In addition, a re-anchored rolling prediction strategy is designed to emulate the receding-horizon mechanism of practical MPC. Validation across multiple wind-speed cases shows that the proposed model can capture short-term turbine dynamics with acceptable accuracy, providing a usable prediction model for subsequent MPC design.

Wind turbine, linear time-varying model, parameter identification, rolling prediction, model predictive control