Model Predictive control (MPC) is a kind of control method with simple principle, multi-variable control and easy to deal with nonlinear constraints, which has been widely studied in power electronics and power transmission fields. In general, the optimal control of the system is realized by minimizing the objective function. However, when the dimensions of control variables in the objective function are inconsistent, the corresponding weight coefficients should be designed to realize the simultaneous control of multiple variables. Taking flux and torque control of induction motor as an example, due to the lack of general theoretical design method to determine the weight coefficient of flux and torque, in order to ensure that the speed regulation system has good dynamic and static performance in different running points, in practical application, a large number of simulation or experiment to determine the size of the weight coefficient. Therefore, an improved MPC for induction motor control is proposed. By analyzing the relationship between flux guide and torque, the simultaneous control of stator flux amplitude and electromagnetic torque is converted to the equivalent control of stator flux vector, thus eliminating the cumbersome weight design in the traditional method, and the algorithm is simple and easy to implement. In addition, the digital control delay is compensated, and the pre-excitation method is used to increase the starting torque and reduce the starting current. Finally, simulation and experiment are carried out on a two-level inverter induction motor platform, and the results show that the proposed method has good performance in a wide speed range.