Research On Direct Torque Control For Induction Motor Based On Neural Network

Direct torque control is the main control method for AC induction motor. Compared with VC control, it reduced the computational complexity of the control system and control the electromagnetic torque directly. But the DTC has the problem that the stator flux is difficult to accurately observe, which makes the fluctuation of the original wave become larger. When the motor speed is lower than the rated value of 30%, this problem is particularly serious. Because of the drawback that the DTC is only close to the rated speed, the effect is good, which severely limits the speed range of DTC control. In spite of the existing schemes, the stator flux linkage observation can be optimized, but it is still a difficult problem to realize the accurate tracking of the stator flux.One of the main reasons stator resistance fluctuation in the low speed region were analyzed.Put forward to design a classifier can be accurately track the stator resistance change, to compensate for resistance to synchronous stator flux observer based on the output results, such changes can ensure synchronous stator resistance flux observer of stator resistance and actual motor. Stator flux observer of stator resistance can get accurate stator flux, the stator flux is involved in the calculation of electromagnetic torque and speed, so the accuracy of flux linkage is improved.Because the neural network has inherent advantages in dealing with nonlinear problems, it is used to realize the real-time observation of stator resistance by using BP neural network. A BP 2-12-1 neural network is designed, and the BP neural network is trained by the traditional gradient descent method. Will train the BP network to be used to track changes in resistance.According to the dynamic mathematical model of the asynchronous motor, the asynchronous motor model and the corresponding DTC system are built in MATLAB 8.1. In order to verify the validity of the design scheme, design a perturbations applied to the motor stator resistance, the first set of experiments only on the motor stator resistance perturbation is applied, flux observer using rated stator resistance; the second set of experiments in the motor stator resistance perturbation is applied by the design scheme from the speed difference in identification of stator resistance variation and the identification results are always to compensate the stator flux observer of stator resistance. The simulation results show that the design scheme can effectively identify the stator resistance variation, and the stator flux observation results can achieve the expected value.