Abstract: To solve the problems of strong nonlinearity of hydrodynamics, and the influences of wind, wave and current on USVs (unmanned surface vehicles), we propose an active enhanced LPV model and yaw keeping controller. The LPV model uses the USV's speed as the varying parameter to simplify system parameters identification and controller design. Based on the proposed LPV model, the disturbances of wind, wave and current are regarded as unmodeled errors estimated online by UKF to improve the overall modeling accuracy. Then, we design a yaw keeping nominal controller based on the proposed LPV model by using linear matrix inequality. Furthermore, we design a feedforward controller to compensate the unmodeled errors online. Simulation results show that the proposed method can effectively reduce the impact of unmodeled errors such as wind, wave and current on the nominal controller.