Research On Nonlinear Modeling And Advanced Control Strategy For Switched Reluctance Motors

Switched reluctance motor(SRM)has the advantages of simple and rugged structure,high efficiency and reliability,flexible control,wide speed range,and low cost,and it is competitive in many fields,including aviation industry,electric vehicle,wind power generation,household appliances,and so on.However,the major drawback of SRMs is that it is difficult to model and control accurately due to its double saliency structure and the discrete commutation from one phase to another.In this dissertation,the key problems including nonlinear modeling,minimizaition the torque ripple,and speed control for the SRM are studied,and the main works can be summarized as follows.First,the problem for modeling and simulation of SRM electromagnetic characteristics based neural network method is studied.The measurement principle and steps of SRM electromagnetic characteristics based on rotor locking method are described in detail,and the flux-linkage and torque characteristics of SRM prototype are measured.Further,an improved back propagation(BP)neural network modeling method is developed to establish accurate models of flux-linkage and torque characteristics of SRM.The method uses the analytical expression which can reflect the nonlinear characteristics of SRM to preprocess the measured sample data.The preprocessed data are used as the additional input of BP neural network to improve the generalization ability of the network and reduce the fitting error of the network.Finally,with the improved neural network,the nonlinear simulation model of a SRM drive system is built in Matalb/Simulink software,and the simulation results have good agreements with those from experiments,which further verify the accuracy of the proposed method.Second,the problem of analytical modeling of SRM electromagnetic characteristics based on small sample flux-linkage characteristics is studied.The torque balanced method is used to measure the flux-linkage curves at four special positions of SRM,and the influences of the magnetic coupling among phases are evaluated in detail.However,only four flux-linkage curves can be obtained,which are inadequate for accurate modeling of SRM.In order to solve this problem,this paper proposes an analytical modeling method based on high-order Fourier series.In the first step,flux-linkage characteristic of SRM in the middle position of an electric period is estimated by the variation characteristics of the flux-linkage with rotor position.In the second step,an analytical model based on five-order Fourier series is developed to construct the entire flux-linkage characteristics,and the analytical expression of torque is derived as well.The validity of the proposed method is further verified by comparing the calculated results with the measured data from the rotor locking method.Third,for the problem of obvious torque ripples of SRMs caused by their double saliency structure and the discrete commutation,this study proposes an improved finite control set predictive torque control(FCS-PTC)algorithm.Firstly,based on the five order Fourier analytical formula,a discrete time predictive model is established for predicting the future states of SRM drive system.Secondly,a new sector partition technique is developed to reduce the candidate switching vectors and computational burden.Further,each prediction is used to evaluate a cost function,and consequently,the vector with minimum cost is selected and generated.The simulation and experiment results demonstrate that the proposed method not only can minimize the torque ripple but also can reduce effectively the copper losses.Fourth,an adaptive radial basis function(RBF)neural network controller is designed for SRM speed control with both parameter variations and external load disturbances.The RBF neural network is employed to approximate an ideal control law which includes parameter variations and external disturbances.The norm of the weights of the neural network is used instead of the weights as the on-line estimation parameters,which reduces the number of the on-line learning parameters from multiple to one.Furthermore,a proportional control term is introduced into the control law,which effectively solves the contradiction between improving the transient performance of the system and chattering of the control law caused by the excessive symbolic function term.The asymptotic stability of the proposed controller is guaranteed through rigorous Lyapunov analysis.The simulations and experiments are carried out to demonstrate the effectiveness of the proposed control scheme.