Abstract: In the current process of handling low-frequency oscillations in power grids, the Power System Stabilizer (PSS) is typically used to suppress the oscillations. However, due to its excessive reliance on precise models, the power amplitude after the suppression processing still remains relatively large in dynamic scenarios. Therefore, a low-frequency oscillation suppression method based on the self-adaptive disturbance rejection control technology is proposed. The Prony algorithm is used to analyze the real-time oscillation signals of the power grid, obtaining parameters such as amplitude and frequency, which serve as the basis for low-frequency oscillation suppression. A self-adaptive disturbance rejection control compensation framework including a tracking differentiator, an expanded state observer, and a nonlinear feedback control law is established. By synchronously inputting the real-time oscillation modal parameters and the expected set values into it, the optimal compensation control quantity can be obtained. Finally, considering the control delay fully, the compensation control quantity is optimized to obtain the best low-frequency oscillation control result. Experimental results show that after this method is applied for suppression processing, the maximum power amplitude of the power grid is reduced from 20 MW to 4 MW, demonstrating its superior performance.