Speed Estimation Methods Of Induction Motor Based On Neural Network

Vector control can solve the problems of the tranditional induction motor drives and makes induction motor easy to control like DC motor, while speed sensorless vector control can eliminate the speed sensors which makes the relative induction motor drives simple, economical and reliable. For this reason, the speed sensorless vector control is becoming the attractive researching project, and under this circumstance, the thesis begins to undertake the study of speed sensorless vector control.Firstly, the thesis introduces the theories of speed sensorless vector control, then selects the neural network (NN) speed estimation scheme after comparing the presented schemes before. The reason is that compared with other traditional control methods, NN control has many advantages, such as self-study, self-adaptability, generalization ability and capability of approaching nonlinear system with arbitrary precision. Besides, in our country, such a research is just being undertaken, so it is necessary to study NN control and apply it to the real systems.Secondly, some NN theories are introduced and three NN speed estimation schemes based on NN mode identification principle are adopted. Also, the relative NN model is established and its structure and study arithmetic is confirmed. Thirdly, a simulation model based on the speed sensorless vector control principle is established, and its good operation performance is proved by simulation. Also, the presented NN speed estimation schemes are discussed and the simulation results show that compared to the traditional schemes, the NN schemes have good speed estimation precision. Moreover, they are robust to the variation of motor parameters and not sensitive to the effect of iron loss.Finally, the good performance of NN speed estimation model is proved once more with experimental data. Besides, by using the NN trained under one condition and the experimental data acquired under another condition to estimate speed, the thesis proves that the NN model can approach the real nonlinear system with arbitrary precision and has the function of fault- tolerance and generalization.