[Objective] Wind turbines operating in turbulent and rapidly varying wind fields are subject to pronounced aerodynamic thrust fluctuations

which propagate into drivetrain and tower dynamics and may accelerate fatigue damage

especially at and above rated wind speed. Conventional baseline controllers for utilityscale turbines typically rely on gain-scheduled PI pitch regulation and region-based torque control. Although robust and widely adopted

such schemes have limited capability to explicitly trade off multiple

coupled objectives(e.g.

tower-top load mitigation and power smoothing) under strongly nonlinear and time-varying aerodynamics. This study aims to develop a computationally efficient control-oriented framework that jointly suppresses tower-top foreaft loads and improves power output quality for a large wind turbine under complex wind conditions. [Methods] A velocity-driven linearization quasi-linear parameter-varying(quasi-LPV) model predictive control(MPC) strategy is proposed. The key idea is to represent the nonlinear turbine dynamics through state and input rate variables

thereby reducing the dependence on stored steady-state vectors across operating points and improving online tractability. To support MPC design

the dominant turbine dynamics are captured via a reduced-order nonlinear model derived from the Lagrange formulation

including the primary rotor/generator rotational dynamics

drivetrain torsion

and tower fore-aft motion. The reduced model is validated against the FAST simulation environment under a stepwise wind profile spanning 4～24 m/s

using time-domain responses of rotor speed

generator speed

aerodynamic thrust

generator power

pitch angle

and tower-top fore-aft acceleration. For control synthesis

wind speed disturbance is explicitly introduced as an augmented state to enhance the prediction of wind fluctuations and to improve disturbance accommodation. A multi-objective MPC is then constructed with coordinated generator torque and collective pitch commands

subject to actuator dynamics(second-order pitch actuator model) and operational constraints. The controller is evaluated on the NREL 5 MW reference wind turbine. Representative load cases are defined according to IEC-oriented settings for the Ijmuiden site

covering mean hub-height wind speeds from near-rated to high wind speed(e.g.

approximately 11.7～20.0 m/s) with associated turbulence intensities. [Results] Model validation shows high consistency between the reduced-order model and FAST in both transient behavior and steady-state gains across below-rated and above-rated regions; the largest steady-state discrepancy occurs near rated wind speed and is most visible in rotor thrust(on the order of several percent)

while key controlrelevant outputs such as generator power and tower-top acceleration remain in acceptable agreement. In closed-loop simulations

the proposed MPC achieves substantial reductions in rotor-speed variability and tower-top fore–aft acceleration compared with the baseline controller

and the improvement becomes more pronounced at higher wind speeds. Across the studied load cases

the standard deviation of rotor speed is reduced by roughly 60% to 90%

and the standard deviation of tower-top fore–aft acceleration is reduced by approximately 20% to nearly 70%. Generator power peaks are also mitigated(on the order of 6%～13%)

lowering overload risk. Near rated conditions

a modest increase in power standard deviation may occur due to the inherent regime transition and the multi-objective weighting that prioritizes structural load alleviation; however

under higher wind speeds

power fluctuations are markedly reduced(tens of percent up to nearly 90%). Moreover

augmenting the wind speed disturbance as a state yields additional benefits in a representative high-wind case

further reducing tower-top acceleration variability and power fluctuation relative to an MPC design without wind disturbance augmentation. [Conclusion] This work demonstrates that a velocity-based linearization quasi-LPV MPC framework

supported by a validated Lagrangebased reduced-order turbine model and an augmented wind-disturbance state

can effectively balance tower-top load suppression and power optimization for a utility-scale wind turbine in complex winds. The controller provides stronger robustness and multi-objective coordination than the baseline strategy

with particularly significant gains in high-wind operating regions. The proposed methodology offers a practical pathway for rapid MPC iteration and for extending verification to other turbine classes and structural configurations in future studies.