Active magnetic bearings feature advantages of frictionlessness, low loss, and high reliability, making them extensively utilized in fields such as flywheel energy storage, aerospace, and beyond. However, conventional modulation strategies applied to digital control systems suffer from control delays, reducing current control precision and resulting in increased current ripple. To address the aforementioned issues, firstly, the operating principle of the active magnetic bearing drive system is analyzed. Based on hybrid systems theory, a mix logical dynamic model of the drive system is established by introducing auxiliary logical variables and auxiliary continuous variables to achieve three-level modulation. Secondly, integrating model predictive control theory, the established model is utilized as a predictive model to forecast and compensate for control delays in controlling current. Finally, a cost function is established based on the error between predicted current and reference current, and optimal control signals are generated to achieve precise control of the active magnetic bearings. The simulation results demonstrate that under light load conditions, the modulation strategy proposed in this paper reduces current ripple by 49.94% compared to traditional modulation strategies. Under moderate load conditions, the proposed modulation strategy reduces current ripple by 49.96%, while under heavy load conditions, it reduces current ripple by 49.99%. This validates the effectiveness of the proposed modulation strategy in compensating for control delays while retaining the three-level modulation scheme.