This paper investigates the motion control problem of Unmanned Surface Vehicles (USVs) during autonomous berthing operations and proposes a control strategy based on Nonlinear Model Predictive Control (NMPC). The approach models the underactuated dynamics of USVs using a three-degrees-of-freedom framework and formulates a constrained optimal control problem incorporating state prediction, control input optimization, and actuator limitations. The NMPC controller, designed for berthing tasks, incorporates nonlinear hydrodynamic effects, input constraints, and disturbance rejection. It further integrates a warm-start strategy and adaptive heading weighting into the cost function to improve robustness and control precision during docking. To validate the proposed method, two representative simulation scenarios are designed, evaluating the impact of different input weighting parameters on control performance. Results show that higher input weights effectively reduce actuator chattering, improve velocity convergence, and significantly enhance berthing accuracy in both position and heading. The findings demonstrate that the NMPC-based approach achieves superior trajectory tracking and robustness compared to conventional control methods, highlighting its strong potential for real-world USV autonomous berthing applications. Furthermore, the results emphasize the importance of controller parameter tuning, providing valuable guidance for future engineering implementation.