Torque-current neural network switch reluctance machine control method and system

Abstract

The invention discloses a torque-current neural network switch reluctance machine control method and system, and the method comprises the steps: configuring one neural network feedforward controller for each phase of an SRM, taking a torque-current inverse model as an activation function, taking the reference torque, distributed through given total torque, of each phase and the rotor position angle as the input, and achieving the feedback error learning through the output of a PID controller. The output of the neural network feedforward controller and the output of the PID controller are superposed to serve as a reference current which is transmitted to a current hysteresis loop controller, so as to control the operation of the SRM through combining with a current feedback signal at a current moment. The system SRM is provided with the current, position and torque sensors, and a signal processor comprises three neural network feedforward controllers, a torque distribution module, a PID control module and a current hysteresis loop control module. An inner loop current hysteresis loop controller tracks a reference current, and controls the operation of the SRM. The method gives full consideration to the special high nonlinearity of the SRM, and effectively reduces the torque pulsation of the SRM.

Description

technical field [0001] The invention relates to the technical field of control of switched reluctance motors for driving new energy vehicles, in particular to a torque-current neural network switched reluctance motor control method and system. Background technique [0002] At present, energy and environmental issues are becoming more and more prominent, and electric vehicles have received extensive attention and have great development prospects. Compared with other motors, SRM (Switched Reluctance Motor, SRM in this article means switched reluctance motor) has many excellent characteristics such as simple and strong structure, no need for rare earth materials, and is suitable for frequent start and stop. The field of electric vehicles has the greatest application potential. However, due to the characteristics of SRM's own double salient pole structure, the magnetic circuit is strongly nonlinear and saturated, which leads to the disadvantages of torque ripple, noise, and lar...

AI Technical Summary

Problems solved by technology

[0005] In the existing literature reports, in SRM modeling and control, the neural network used is a general neural network model, and the neural network model design is not combined with the special nonlinear characteristics of SRM, so it is difficult to effectively control the SRM torque

Images