A microgrid is an autonomous active distribution network system that integrates distributed new energy sources, loads, energy storage, as well as monitoring and protection devices. The servo motor system is a key executive and control unit in the microgrid, and its operation accuracy and stability are directly related to the flexible control capability of the microgrid. As the mainstream executive mechanism of the servo system, permanent magnet synchronous motor (PMSM) is easily affected by various uncertain factors such as parameter mismatch and unknown load disturbance during actual operation. However, the traditional current control strategy has the problems of low current tracking accuracy and insufficient operation stability. Hence, this paper proposes a disturbance compensation method for deadbeat predictive current control based on an improved sliding mode observer. Firstly, an exponential-power composite approaching law is employed to suppress the chattering phenomenon inherent in traditional sliding mode control, enabling precise observation of comprehensive disturbances including internal parameter mismatches and external load disturbances. Secondly, combining the mathematical model of the PMSM with the principle of deadbeat predictive control, the observed disturbance quantity from the observer is integrated into the deadbeat predictive current model. This constructs an optimal dq-axis voltage reference value solving mechanism that incorporates a disturbance compensation term, thereby forming a deadbeat predictive current controller integrated with a sliding mode observer. Finally, strict stability analysis is conducted based on Lyapunov stability theory, and the stability of the observer and the closed-loop system is proven. The effectiveness of the proposed method is verified through simulations and experiments.