Abstract: To address the prominent wind induced vibration of ultra large deep sea wind turbines and the limited energy supply for active control, this study proposes a parameter optimization method for a virtual tuned mass damper (VTMD) that integrates vibration mitigation and energy harvesting. A Gaussian process regression (GPR) surrogate model is introduced to improve the computational efficiency of the optimization process. First, an electromechanically coupled VTMD model for the IEA 15 MW wind turbine is established, and the mapping relationship between the control force and battery power is formulated. Second, the virtual mass ratio, tuning frequency, and damping ratio are selected as design variables, while the root-mean-square value of tower-top acceleration, peak acceleration, and average battery power are adopted as evaluation indicators. Based on 500 samples, a GPR surrogate model is constructed to perform parameter optimization. Finally, closed-loop simulation verification is performed under turbulent wind conditions with a mean wind speed of 16 m·s⁻¹ at the hub height. The results show that the established GPR surrogate model achieves high prediction accuracy and enables rapid parameter optimization. The optimized VTMD exhibits better vibration mitigation performance than the passive tuned mass damper, and the energy-harvesting-oriented parameter combination achieves energy recovery while maintaining vibration control capability.