Research On Adaptive Vector Control System Based On Neural Network

Field-oriented control has proved to be one of the most significant methods to achieve high performance in adjustable speed induction motor drives. The basic task of this scheme is the decoupling of the rotor flux terms, which permits torque attributes similar to those of dc machines. Decoupling is accomplished by the application of a slip command signal dependent on the estimate of the motor's parameters. Accordingly, the commanded stator currents decomposed into the flux and torque producing components, which are independently controlled. The realization of the field-oriented control scheme requires exact knowledge of the machine's parameters. However, some of these parameters cannot be directly measured. Furthermore, the major disadvantage of the field-oriented control is its sensitivity to rotor time constant variation. Inaccurate estimate of this parameter will cause the deterioration of the performance of the induction motor drive. Therefore, an observer is required to provide the parameters estimate.On the other hand, it has been recognized that a controller based on neural network architecture offers a promising way of handling complex control problem. Neural network have been used successfully for a wide variety of application such as pattern recognition, vision and speech processing, control, and identification.Influence on field-oriented control system by the motor's parameters' variations is analyzed in this paper. And the equation between the parameters and the instantaneous reactive power of the induction motor in the field-oriented control system is deduced. With the equation discussed above, the task of parameter identification and self-adapted is performed by a neural network based adaptive observer. Extensive simulations have shown that this scheme works well in case of motor control.