Active Disturbance Rejection Control For An Axial-flux Permanent-magnet Machine Drive System

Compared with traditional radial-flux permanent-magnet machines,axial-flux permanentmagnet machines(AFPMMs)have been widely studied and applied in the field of electric vehicles due to their compact space and high power density.However,disturbances includin g motor flux harmonics,cogging torque,inverter nonlinearity,parameter changes and external interference will seriously affect the control performance of the motor system,especially in the robustness and anti-disturbance performance of the control system.The aspect urgently needs to be improved.Active disturbance rejection control(ADRC)is the equivalent of the controlle d object as an integral series model,and all other information is defined as a centralize d disturbance through feedforward compensation.It is widely used due to its strong universality.From the perspective of active disturbance rejection control,the article aims to improve the speed and torque control performance of AFPMMs,and conducts special researches on speed ripple suppression algorithms,torque ripple suppression,and speed detection algorithms.To solve the problem of difficulty in parameter tuning of nonlinear ADRC,an ADRC controller of AFPMM based on back propagation(BP)neural network and multi-objec t ive evolutionary algorithm based on decomposition(MOEA/D)is proposed.Among them,the parameter tuning method based on BP neural network can calculate the current control parameters in real time through the backpropagation algorithm.Online parameter tuning reduces the controller’s dependence on parameters;for applications with multiple time doma in index requirements,use Multi-objective optimization establishes a multi-leve l evaluation inde x system for the speed,stability and accuracy of the motor drive system to improve the control performance of the motor control system.It focuses on the establishment of the objective function and the process of parameter optimization.The control effect of the intellige nt algorithm and the traditional method is compared through experiments,and the effectiveness of the automatic disturbance rejection control strategy of the axial flux motor based on the intellige nt algorithm is highlighted.Aiming at the problem of AFPMM torque pulsation caused by periodic disturbance,a torque ripple suppression strategy based on active disturbance rejection based iterative learning control(ADR-ILC)is proposed.Firstly,it analyzes the stability analysis of the ADR-ILC for an AFPMM control system based on the logarithmic stability criterion,and derives the feasible range of the controller parameters according to the system stability margin and the ma pping relationship selected by the parameters,and then designs The parameter adaptation rate of the iterative learning module is improved,the adjustment time is shortened,and the problem of the dynamic performance degradation of the traditional iterative learning control system is solved.The experimental results show that the ADR-ILC proposed in this paper can reduce nearly half of the torque ripple in steady state,and compared with the fixed parameter iterative learning control strategy,it shortens the system dynamic response adjustment time.Improve the dynamic performance of the system.Due to the high cost of the traditional resolver dedicated chip for decoding,and does not include the preprocessing of the sine and cosine winding signals,a soft decoding solution based on resolver is proposed.The least square curve fitting method is used to compensate the zero offset error and amplitude error,and reduce the periodic disturbance of the system.At the same time,the idea of active disturbance rejection is used to estimate the concentrated disturbanc e,and an angle observation method based on the extended state observer(ESO)is proposed.The experiment verifies that the algorithm has a higher dynamic decoding accuracy than the secondorder angle observer.This paper takes the AFPMM of model EMRAX228 as the control plant,and builds the d SPACE Micro Lab Box real-time simulation hardware experiment platform with Matlab/Simulink as the core.The research of torque ripple suppression by auto-disturbance-rejec tio n iterative learning and the resolver soft decoding control strategy based on the extended state observer are simulated and experimentally verified.The results show that the speed and torque control performance of the axial flux motor drive system have been effectively improved in all aspects.