In order to realize the quantitative control of the flux and torque in permanent magnet synchronous motor (PMSM), based on the simplified models of flux and torque variation, the deadbeat (DB) control for PMSM is proposed. The torque angle is important to DB control. It can be obtained through look\|up table or the d and q components of stator flux. However, the former needs lots of memory space and the latter needs rotor flux position information. To solve these problems, the back propagation (BP) neural network is used to predict torque angle. Another BP neural network is used to predict the angle of ideal voltage vector to replace conventional calculation. Thus, the DB control of stator flux and torque for PMSM with dual BP neural networks is established. Simulation results show that the BP neural networks can be used to output torque angle and the angle of ideal voltage vector and the motor system works properly.