Research On Torque Observer Of Switched Reluctance Motor Based On Neural Network

Environmental pollution and energy shortage are the main problems social development now faces with.Switched reluctance motor(SRM),which is energy-efficient and environmental,has become a research hotspot in the field of motors.SRM has many advantages such as simple structure,low cost and high efficiency,which makes it have broad application prospects in socio-economic development.Torque is the main control variable of SRM and its accurate calculation is of great significance.However,due to the inherent characteristics of SRM,such as doubly salient structure,nonlinear magnetic circuit and high magnetic saturation,torque calculation is much more complicated.It makes the modeling of SRM difficult and also results in large torque ripple.Undoubtedly it limits the application of SRM.Torque estimation methods commonly used in SRM are analyzed in this paper.Aiming at the shortcomings of SRM torque observer based on traditional BP neural network(BPNN),this paper proposes a new SRM torque observer based on BPNN optimized by Bat Algorithm(BA).In this method,the advantages of BA algorithm in finding the global optimal solution was used to optimize the initial weight and threshold of BPNN so that it can improve the convergence speed and prediction.accuracy of BPNN.Subsequently,the nonlinear relationship among SRM current,rotor position angle and torque was mapped by BA-BPNN.And a SRM torque observer based on BA-BP neural network was constructed.After constructing the torque observer,it was first applied to SRM nonlinear modeling,and the SRM ontology model was constructed together with current look-up table module.Secondly,to verify that a more accurate torque value is helpful to further reducing torque ripple,it was applied to control SRM with the direct torque control method.Corresponding simulation systems were built on MATLAB/Simulink to verify the feasibility of the proposed torque observer application.Simulation results show that the proposed modeling method is feasible and the torque control method can effectively minimize the torque ripple of SRM.Finally,a SRM experimental platform was built to verify the application of BA-BPNN torque observer.Through the comparison and analysis of experimental waveforms,the correctness and feasibility of the proposed torque observer method in SRM nonlinear modeling and torque ripple suppression are verified.