There is decreased control performance of induction motors employing the traditional proportional integral (PI) control algorithm when facing complex disturbances. Thus a model-free predictive control (MFPC) method based on active disturbance rejection control (ADRC) is proposed using a vector control system for induction motors. First, combining the mathematical models of the speed and magnetic chain loops, ADRC controllers for those loops are designed to observe and compensate for both internal and external disturbances caused by load variations and internal parameter perturbations. Secondly, to avoid the dependence of the inner loop controller on motor parameters, a hyperlocal equation for the dq current loop is established based on the model-free control principle. Variables, except the control variables, are treated as disturbances, and a nonlinear extended state observer is introduced to estimate them. Finally, incorporating predictive control principles, a current loop controller is designed to obtain the switching state signals applied to the inverter. Simulation and experimental results demonstrate that the proposed algorithm exhibits better immunity and robustness than the PI algorithm, and can effectively improve the dynamic and steady state performance of the induction motor.