Neural Network Inductance Model Construction And Torque Torque Ripple Suppression For Switched Reluctance Motor

Switched reluctance motor is a new type of special vehicle motor.Its advantages are high reliability,low manufacturing and maintenance cost,wide speed regulation range and good heat dissipation performance.These advantages make it widely used in industrial production,machinery manufacturing and automotive electronics.However,the high saturation and strong nonlinearity of the magnetic circuit of switched reluctance motor led to large torque ripple and poor aging performance at low speed.However,the high saturation and strong nonlinearity of the magnetic circuit of the SRM lead to large torque ripple,resulting in poor performance at low speed.In order to reduce the adverse effects of torque ripple,two control strategies for SRM torque ripple suppression are proposed in this article.(1)Inductance model construction and torque ripple suppression control of D-Sigmoid neural network based on mechanism characteristicsDifferent from other automotive motors such as permanent magnet synchronous motors,the strong nonlinear of SRM is determined by the inductance nonlinear mechanism.According to the mechanism characteristics of SRM phase inductance,the D-Sigmoid neural network excitation function is designed to construct the neural network inductance model.During the model learning process,the torque deviation preprocessing method is introduced to accelerate the parameter correction of the neural network inductance model.Based on the constructed neural network inductance model,the torque-flux linkage model is designed,its output is used as the reference flux linkage for the inner loop control of the flux linkage,thus realizing the effective control of SRM torque ripple.The neural network control strategy with D-Sigmoid as the excitation function fully described the inductive nonlinear mechanism characteristics of the SRM,improved the dynamic performance of the system and suppressed the torque pulsation.(2)SRM torque ripple suppression control based on Fuzzy fractional PID and Dsigmoid neural networkFractional-order differentiation is used to improve the dynamic characteristics of the system and weaken the external high-frequency interference introduced by the system;Fractional-order integral has the characteristics of anti-integration saturation,which reduces the steady-state error of the system and effectively avoids the phenomenon of integration saturation.The fuzzy control fuzzified the fractional differential output,which further overcomes the external disturbance of the system;The D-Sigmoid neural network models the total SRM inductance and the Torque Sharing Function(TSF)distributes the total inductance.Compared with the mechanistic-based D-Sigmoid neural network inductance modeling and torque pulsation suppression control strategy,it can accurately describe the nonlinear mechanistic characteristics of the phase inductance and reduce the computational effort of neural network parameter correction.The simulation results show that the SRM flux linkage control strategy with fuzzy fractional-order PID incorporating D-Sigmoid neural network can effectively reduce the torque pulsation at low-speed operation of SRM.On the SRM control system experimental platform with the model STM32F407ZGT6 chip as the core,the D-Sigmoid neural network inductance model construction and torque ripple suppression control strategy proposed in this paper based on mechanism characteristics are tested.The experimental results verify that the D-Sigmoid neural network inductance model construction and torque ripple suppression control strategy based on the mechanism characteristics can effectively suppress the torque ripple of the SRM and have good dynamic characteristics.