Active Disturbance Rejection Control of Bearingless Permanent Magnet Synchronous Motor Based on Genetic Algorithm and Neural Network Parameters Dynamic Adjustment Method

Abstract

In order to solve the problem of poor control performance, caused by fixed parameters of the active disturbance rejection control (ADRC) in bearingless permanent magnet synchronous motors (BPMSM), a dynamic parameters adjustment method of ADRC, based on a genetic algorithm and back-propagation neural network (GA-BPNN), is proposed. Firstly, the ADRC control models of motor-side and suspension-side are established, according to the motor speed formula and suspension force formula. Secondly, the BPNN algorithm is used to dynamically adjust the parameters of the ADRC, and the operation processes of BPNN are deduced, according to the chain rule. Thirdly, in order to avoid the problem of getting out of control, caused by the convergence failure of BPNN, a GA based on floating point coding is used to optimize the initial value of BPNN. Finally, these methods are integrated to form a BPMSM control system, based on the GA-BPNN-ADRC, and the effectiveness is verified on an experimental platform. The experimental results, show that the proposed method reduces the failure probability of the system from 35.61% to 0%, and the anti-interference ability and dynamic performance of the speed and displacement of the control system are significantly improved.