The drivetrain of high-power wind turbines has a large moment of inertia, resulting in a high risk of low-frequency torsional vibration when subjected to random wind speeds. The traditional torsional vibration control method for the drivetrain based on the determination of the torsional speed control target fails to consider the influence of the random measurement noise introduced by the measurement equipment on the torsional speed. This oversight can lead to a decrease in control performance. To address the uncertainty associated with torsion speed measurement, this paper proposed an active disturbance rejection control method for torsional vibration of large-scale wind turbine drivetrain. A KF-ADRC torsional vibration controller was designed which dynamically estimated the torsion speed of drivetrain using a Kalman filter. The ADRC control target was set as zero torsion speed to regulate the electromagnetic torque of the generator and suppress low-frequency torsional vibration in the drivetrain. The research results show that when the input signal of the conventional torsional vibration controller has random measurement noise, it will significantly reduce its performance. In contrast, the KF-ADRC torsional vibration controller can effectively predict the drivetrain’s torsional speed and achieve superior suppression of the low-frequency torsional vibration in the drivetrain.