This paper proposes a data-driven Model Predictive Control (MPC) for advanced wind turbine (WT) control considering the mechanical load mitigation. Specifically, we use a neural network applying rectified linear unit (ReLU) as the activation function for the hidden layers to capture nonlinear WT dynamics under all wind conditions so as to facilitate rotor speed regulation and suppress fore-aft tower bending. Case studies demonstrate that the approach has better damping for shaft torque oscillation and tower for-aft bending than the baseline control. This marks a notable enhancement in the overall stability and structural robustness of the WT system, ultimately resulting in increased efficiency and sustainability in its operation.