Research On Flux Identification Based On Neural Network Of Direct Torque Control System For Induction Motor

Direct Torque Control (DTC) is after vector control a novel high-performance technologies AC speed drive control, abandoned the decoupling control of vector control, largely solved some technical problems, such as computational complexity of vector control、features vulnerable to the impact of motor parameters, because of novel control、simple method, advantageous of static and dynamic characteristics, is the AC drive one of the hot areas of research.Double-flux and torque hysteresis control are used in the conventional DTC system based on looking up switching vector table. There are some problems such as large ripples of the flux and the torque and current、variable switching frequency. In the essay, a flux Identification of induction motor DTC system (SVM-DTC) based on neural network is proposed, based on the traditional DTC system using space vector pulse width modulation theory.SVM-DTC control method based on the error of torque and stator flux using space vector modulation theory, real-time synthesize an optimal voltage vector acting on the motor, to compensate the error of motor torque and stator, and reduce ripples of the motor torque and flux; using the symmetry vector modulation can keep switching frequency constant and current sinusoidal degree good, reduce the motor running noise.In order to improve the control performance of DTC system, many scholars are working on motor parameter identification technique of high precision about rotor resistance, speed, rotor and stator flux etc. Because the DTC system achieves the control of the electromagnetic torque by controlling the motor stator flux, the accurate identification of stator flux is the key to ensure the performance of DTC control system, so this paper mainly studies the identification of stator flux of DTC system. Based on the comparison between traditional open-loop stator flux identification methods, the stator flux identification method based on adaptive integrator and the closed-loop stator flux identification method based on extended Kalman filter, this paper proposed a stator flux identification method based on neural network. This identification method using BP neural network, the neural network’s input signal has dynamic filtering processing. The method has many advantages such as simple learning algorithm, good stability and high precision and etc, can solve the problem of initial value and integral saturation of pure integrator, and the problem of low-pass filter amplitude and phase shifting, and has good robustness on the stator resistance.In this paper, on the platforms of MatlabR7.8/Simulink, the stator flux identification based on neural network of induction motor SVM-DTC system is simulate, the simulation results are in accordance with the theoretical analysis. Finally, based on dspic6010A the hardware and software for induction motor direct torque control system are designed, and the experimental results properly show direct torque fast dynamic response, and prove the identification method is correct.