Abstract: Cogging torque in permanent magnet synchronous motors leads to speed fluctuation and deterioration of control accuracy. To address the strong parameter dependency of traditional compensation methods, a suppression scheme based on disturbance observer is proposed. By constructing a nonlinear observer to estimate cogging torque in real-time and feedforward it to the current loop, dynamic compensation is achieved. Experimental results demonstrate that this strategy reduces speed fluctuation by 68%, decreases torque ripple by 74%, and shortens dynamic response time to 42% of traditional methods, validating the effectiveness of the proposed scheme.