| Switched reluctance drive system with excellent speed performance,simple structure,reliable operation characteristics,and many other advantages,has caused extensive concern of the domestic and foreign scholars.The Direct Torque Control(DTC)has been applied in SRD system because of its advantages such as simple system,strong robustness,good static performance and dynamic response performance.But DTC technology still has some shortcomings that need improvement.Such as large torque ripple,inaccurate flux linkage observation,and unstable switching frequency,which affect the overall performance of the system.In this paper,we mainly study the stator flux observer and the torque observer.In view of the shortcomings of the flux observer and the torque observer,a new method of flux observation and torque observation are proposed.Aiming at the shortcomings of the flux observer and the torque observer in the traditional DTC,the finite element analysis and BP neural network technique were used to establish the flux observer and the torque observer.The first method uses the finite element analysis software to model the switched reluctance motor,using the modeling of the flux linkage curve and the torque characteristic curve to establish the flux observer and the torque observer.The second method is to establish the flux observer using BP neural network,in order to improve the accuracy of BP neural network,particle swarm optimization algorithm is used to optimize BP neural network,the BP neural network flux observer based on particle swarm optimization is established.On the MATLAB simulation platform,firstly,the flux observer and torque observer based on finite element analysis are simulated and analyzed,secondly,the flux observer based on BP neural network is simulated and analyzed.The simulation results show that the two methods do improve the accuracy of the flux observer and the torque observer.Finally,the flux observers established by these two methods are compared and laid the foundation for the follow-up study of the two methods. |