The Application Of Dynamic Recurrent Neural For Induction Motor Control

Due to multivariable, highly nonlinear, strong coupling, time-varying dynamics and unavailability of measurements, induction motor control is still a difficult and complex engineering problem. Vector control has replaced traditional control method using the ratio of voltage and frequency as a constant, which improve greatly dynamic control efficiency of motor. However, under the circumstances of changing of motor parameters, exterior load disturbance and without model of object, the motor performances can be affected. In this paper, first, the configuration and algorithm of a kind of recurrent network are researched. Second, a vector control method of neural network PID combined with recurrent network identifier has been designed, which make use of the excellent performance of recurrent network at nonlinear and uncertain system control and identification. Consequently, control system is proved to be strong adaptive and disturbance rejection through simulation experiments. The paper is combined with two parts as follows:In the fist part, the development of neural network and model of asynchronous machine are briefly introduced. At the same time a recurrent network adaptive control is presented based on stator magnet-oriented control scheme of asynchronous machine. After the simulation of two methods, the results are analyzed, in which recurrent network system can be easily used to improve system performance in an adaptive control of nonlinear dynamic motor model.In the second part, the software and hardware design of a motor control system based on the TMS320f240 is presented. Then the design of power inverter is specified in the article including the drive circuit of power module, design of three-phase input switching power and detection circuit of current, voltage and photoelectric pulse encoder. Finally the problems of testing and the performance of the hardware are analyzed in detail.