Modeling, Torque Estimation And Ripple Suppression For Switched Reluctance Motors

As the preferred motor for the next generation of new energy vehicle drive motors,the switched reluctance motor(SRM)rotor does not require magnetic material.It operates independently of each phase,with high reliability.However,its doubly salient structure makes the inductance highly nonlinear.Therefore,torque ripple is worse at low speeds if the control method for other motors with weak nonlinear properties is transplanted and utilized as the SRM control method.The vibration caused by high torque ripple will reduce the lifespan of the vehicle.It is usual to achieve the objective of producing a stable torque indirectly by managing the current or flux linkage;however,there are disadvantages,such as the inability of parameters to self-learn or the requirement for offline modeling and learning.The mechanism relationship and properties of the torque current flux linkage in SRM establish the foundation for this research work.Research has been done on torque estimation,torque ripple suppression,and intelligent modeling of torque and current in switching reluctance motors.The thorough research includes:(1)A control approach for SRM torque ripple reduction was developed based on fuzzy nonlinear current compensation.The existing PD compensation algorithm obtains torque error by introducing instantaneous torque,generates compensation current through the PD algorithm,and combines it with the torque allocation method to form a currentcontrolled SRM torque ripple suppression method.Based on this,a fuzzy rule table is designed with the mechanical characteristics between SRM current and torque and input and output quantization factors designed according to the torque ripple feature.Fuzzy nonlinear current compensation control is achieved by swapping out the PD compensation algorithm for a fuzzy rule table,which lowers the torque ripple rate by61%.(2)A total current neural network-based control approach for SRM torque ripple suppression was proposed.Based on the nonlinear relationship between current and output torque,the indirect torque control method regulates output torque by modulating current.In this thesis,the current neural network model is built to achieve online learning for the current model.Its implicit function is produced following the switching reluctance motor’s phase current form,which can narrow the optimization range and quicken model parameter learning.Furthermore,its control method has the anti-interference ability and can adapt to situations with current noise interference,reducing its torque ripple rate by68%.(3)A control method for SRM torque ripple suppression based on reference torque comprising PI control and a neural network in series is proposed.There is an error between the ideal reference torque and the reference torque output by the speed PI controller of the indirect torque control method.Therefore,a neural network is connected in series after PI control to form a reference torque model.According to the periodic variation of the reference torque with the rotor angle,the rotor angle is used as the input signal of the reference torque neural network(RTNN)to describe the characteristics of the periodic variation of the reference torque.The RTNN implements online learning based on torque error.The designed control method achieves a 71% reduction in torque ripple rate.(4)A torque estimation method was designed based on a nonlinear inductance neural network model.Calculating and estimating instantaneous torque instead of utilizing torque sensors is an essential technical approach to improve system reliability.The estimation of nonlinear inductance is the key to torque estimation.Based on the inductance mechanism characteristics,the position angle and current are selected as the input variables of the neural network,and the nonlinear inductance neural network model is established.Then,estimate the instantaneous torque through the relationship between SRM torque and the inductance partial derivative.In the SRM control system based on the torque sharing function(TSF),the estimated instantaneous torque is employed in TSF control method.Compared with the linearized torque allocation method,the torque ripple rate is reduced by 78%.(5)An experimental platform for the SRM was designed and developed using STM32F407 as the control unit.The system mainly includes functional modules such as a power conversion circuit,signal isolation circuit,driving methodcircuit,various signal detection circuits,auxiliary power supply circuit,and a logic processing circuit.Based on the established SRM experimental platform,experimental tests were conducted at different speeds and load torques to demonstrate the effectiveness of the proposed method in this thesis.Furthermore,the proposed fuzzy current compensation method,neural network online current estimation method,and neural network bond with PI to estimate reference torque method are compared with the PD compensation method regarding torque ripple.As verified by experiments,the introduction of immediate torque,the utilization of torque error to generate compensatory current,the employment of torque error for current modeling,and the application of torque error for reference torque modeling can all be used to optimize torque performance and suppress torque ripple.The different control methods proposed based on current,torque,and inductance have different requirements for the computing power of the hardware platform,can be employed as a control method individually,or they can be combined to further improve the effectiveness of suppressing torque ripple in SRM and meet the particular needs of high-precision control in SRM.