Abstract: To optimize the overall loss and enhance the parameter perturbation adaptability of permanent magnet wind power systems, this paper first analyzes the motor loss model considering iron loss and the converter loss model, deriving the constraint conditions for both motor and converter losses. Secondly, an ultra-local model is established, and its parameters are estimated using a sliding mode observer and Kalman filter to improve the model's adaptability to parameter variations. Then, a cost function incorporating tracking performance, motor loss, and converter switching frequency is constructed to achieve a balance between control performance and loss reduction. Finally, the effectiveness of the proposed method is verified based on the Simulink platform. The results demonstrate that compared with the traditional maximum torque per ampere (MTPA) control, the proposed control method reduces the loss by 11% at a single operating point; compared with the traditional space vector pulse width modulation (SVPWM), the converter switching frequency is decreased by 40%. Significant optimization effects on both loss and switching frequency are achieved across different operating points. Meanwhile, under parameter perturbations, the control performance outperforms the traditional Proportional-Integral (PI) control and model predictive control, validating the effectiveness of the proposed method.