Abstract: Model predictive control in voltage source PWM rectifier has been extensively studied for many years, which has several advantages such as simple principle, fast dynamic response and easy implementation. However, it poses high computational burden and requires the accurate system model and parameters. By analyzing the mathematical model of PWM rectifier, and combining prediction control and fast vector selection, a direct current control (DCC) is obtained. This method can select the optimal vector through only one prediction and comparison. Then the proposed DCC can be simplified into a new model-free predictive current control (MFPCC), which is characterized by sampling the grid current only once in each control cycle and is independent of the rectifier model and parameters. This method has strong robustness and good dynamic and steady state performance. A comparative study between DCC and MFPCC is carried out in terms of steady state performance, dynamic response and robustness. Finally, the correctness and effectiveness of those methods is experimentally validated.