Direct Torque Control Method Of SRM Based On Neural Network

SRM has low cost, strong, wide speed range, good performance and high reliability. In recent years, it has been paid widespread attention. However, its special structure with big noise and big torque ripple is still a problem. This paper starts from the control point of view to explore the improvement of its torque ripple and noise.This paper study a four-phase SRM which structure is 8/6. Analysis of direct torque control of the space voltage vector and the corresponding switching table.A reasonable choice of the SVPWM will limit the torque ripple in a small hysteresis error band. Therefore, the trajectory of the flux becomes a circular, and stator current and flux change in a smooth waveform. The simulation results can be seen from the direct torque control and effective control of the SRM torque ripple, thereby reducing the electrical noise. In order to have better control of motor speed, the direct torque control system uses PI speed controller to achieve a good speed effect.This paper introduces the model of ANN; learning rules for a typical neural network, study control strategy of basic principles and algorithms for single neural network ;BP and RBF neural network adaptive PID, and used this control strategy in Switched Reluctance Motor Drive. Simulation results show that the combined controller of neural network adaptive PID and Normal PI has good control performance, better robustness and adaptive ability and other advantages; it can achieve the purpose of control with high control accuracy and good dynamic characteristics.This paper describes TMS320F2812 DSP as the core controller. compile the current chopper procedures, and directed torque procedures and other relevant subprogram; it has realized the motor current chopper start and has laid the foundation for further research.