Research On Nonlinear Modeling And Control Of Inductance For Switched Reluctance Motor

Switched Reluctance Motor(SRM)has the advantages of simple and sturdy structure,low manufacturing cost,high reliability of system,and a wide range of speed adjustments,which makes it an important candidate for a new generation of new energy vehicle driving devices.However,the development and application of SRM are severely restricted in the field of new energy electric vehicle by the large torque ripple and noise caused by it.Therefore,in order to effectively suppress the torque ripple and realize the constant torque control of SRM,two control strategies are proposed in the paper.(1)Neural network modeling and control strategy for nonlinear characteristics of SRM inductance change rate.According to the functional relationship between the control current required for constant torque and inductance change rate,this strategy uses neural network to model the inductance change rate,and the specific process is shown as follows: the segmented excitation function is designed according to the characteristic curve of inductance change rate for SRM,and the neural network is built based on it.Since the constructed excitation function describes the actual mechanism characteristic of inductance change rate for SRM,it can effectively improve the modeling speed and accuracy,and realize accurate modeling for the nonlinear inductance change rate.Different from the traditional nonlinear modeling strategy for inductance characteristics,it adopts the direct modeling method for inductance change rate,so that the control current can be obtained more directly and simply.Simulation results show that the proposed SRM control strategy can better adapt to the nonlinear characteristics of inductance change rate,improve the dynamic quality of the system,and effectively suppress the torque ripple for SRM.(2)Inductance model nonlinear compensation and control strategy for SRM based on integrating reinforcement learning and fuzzy inference.It adopts a method of feedforward compensation,which avoids the direct modeling for nonlinear inductance.Fuzzy inference rules are designed and fuzzy compensator is constructed according to torque ripple characteristics of SRM.In the closed-loop control system based on the linear inductance model,the nonlinear feedforward compensation of the linear inductance model is realized by using the torque error and its derivative which can reflect model error of SRM,thus indirectly describing the strong nonlinear characteristics of inductance.Introducing reinforcement learning and designing reward function,the nonlinear adaptive optimization compensation for linear inductance model is further realized by cooperation with fuzzy compensator.The simulation results show that the proposed SRM control strategy can better adapt to the nonlinear characteristics of inductance,improve the dynamic quality of the system,and effectively suppress the torque ripple for SRM.The proposed neural network modeling and control strategy for the nonlinear characteristics of SRM inductance change rate is experimentally verified on an SRM experimental platform with Infineon XC2765X-104 F control board as the core.The experimental results show that compared with the traditional current sharing control strategy for SRM based on linear inductance model,the proposed control strategy can effectively suppress torque ripple of the system and have better performance.