In view of the problem of a poor frequency response and transient stability during the power grid voltage drop and recovery found in traditional VSG constant active power and constant reactive power control strategies under the condition of an unbalanced grid voltage, an MPC-VSG time domain optimal control strategy has thus been proposed for direct driven permanent magnet synchronous wind power generation (D-PMSWG) system under an unbalanced grid voltage. With virtual synchronous machine control (VSG) under an unbalanced grid voltage combined with model predictive control (MPC), by optimizing the prediction time domain, the coefficients of frequency link, damping link and feedback control inertia link are adjusted to obtain the optimal prediction time domain output vector and the control input vector at the next moment. Compared with traditional methods, the proposed method is characterized with an improvement of the frequency response and power transient characteristics at the moment of the voltage drop and recovery. Finally, the accuracy and validity of the control strategy under different control objectives can be verified in Matlab/Simulink.