Three-phase voltage-source grid-connected inverters are widely applied in the field of photovoltaic power generation. A variety of control strategies about inverters have been proposed in order to improve performance and efficiency of grid-connected systems. Model predictive control strategy uses the discrete time model to predict all the possible output value in the next sampling period. The optimal voltage vector is chosen according to the cost function. A model predictive control strategy is used in three-phase voltage-source grid-connected inverter. Firstly, instantaneous power mathematical model of three-phase PV inverter in d-q coordinates systems is established. Secondly, predictive function is designed to predict the inverter grid parameters online. The optimal objective function is selected to control the inverter in the next cycle. The tracking in d-q coordinate system is fast and accurate. The calculation amount of proposed strategy is small. It is easy to be implemented and does not need PWM module. Then, multi-objective optimization control is designed. Current decouple control is designed to reduce reactive power of the system. Cost function is changed to improve output current quality. The AC voltages are modified to improve the prediction accuracy. Finally, simulation and experimental results show that the output current of the proposed strategy has good static performance and low harmonic distortion rate. It can track reference and realize reactive power compensation.