Abstract: The BP neural network was used to predict the speed in the three-phase induction motor vector control system. The training method of BP neural network was improved,the input parameters of motor's stator voltage and current were replaced after analyzing its mathematical model. A new heuristic algorithm was proposed to preprocess the input parameters due to the mismatch between the input and output frequencies of the neural network. Simulation results showed that the improved method had a good tracking performance and static & dynamic performance with a smaller overshoot.