Research On Control Technique Of Weight-transducerless Elevator Permanent Traction Machine Based Onmodel Predictive Control

Permanent magnet synchronous traction machines, due to high torque at low speed, high power density, no mechanical room and many other advantages, are increasingly used in applications of the elevator drive system. However, if the electromagnetic torque generated by the traction machine cannot timely balance the intense changed load torque, the traction sheave will slide instantly. If it happened, the elevator car would slide as long distance as the traction sheave because of no gear transmission. Additionally, the weight transducer is susceptible to be interfered by noise signals, which can lead to incorrect torque compensation and reduce the robustness of the whole control system. However, low-resolution incremental encoders are still widely used in elevator drive systems as feedback devices due to the low cost. Therefore, this thesis mainly focuses on the rollback mitigation of direct-drive permanent magnet synchronous traction machines without the car load feedback during the elevator startup, when using a low-resolution incremental encoder as the position feedback device.In this paper, based on the development analysis of permanent traction machines, starting torque compensation and model predictive control, the mathematical model of direct-drive elevator traction machine and the torque characteristics of traction sheave are established. According to the nonlinear, uncertain and intense changed torque characteristics, the necessity and superiority of newly designed predictive control strategy to mitigate rollback during the elevator startup have been demonstrated.A novel anti-rollback starting torque control strategy based on off-set free model predictive control is proposed to reduce the sliding distance and improve the elevator riding comfort. The model corrector is added to overcome the mismatch of the predictive model caused by the disturbance torque exerted on the sheave and can eliminate the steady-state speed error. The estimated speed is used as the initial predictive speed, so there is no need to correct the predictive speeds in the cost function, which can effectively reduce the amount of computation. Additionally, the stability analysis is carried out and the parameter selection is illustrated.In order to further improve the performance of traction system, an anti-rollback starting torque control strategy based on enhanced model predictive control with extended state observer and nonlinear tracking differentiator is proposed. Extended state observer is used to eliminate the mismatch error caused by the uncertain generalized disturbance. And the nonlinear tracking differentiator can provide smooth and precise speed feedback with less time lag, which can enhance the performance of extended state observer. The stability analysis of the extended state observer is presented. And the parameter selection of both the extended state observer and the nonlinear tracking differentiator is illustrated.Finally, the two newly designed rollback mitigation strategies are validated on both Matlab/Simulink simulation platform and the 11.7k W permanent magnet synchronous traction machine load test platform. By adopting an ordinary 2048-line incremental optical encoder and without car weight feedback information, the experimental comparisons between the two kinds of proposed control strategies and PI controller under different loads are carried out, respectively. The experimental results validates that the proposed rollback mitigation control strategies can achieve better anti-rollback performance during the elevator startup. According to experimental results, contrastive analysis of the two proposed methods has been illustrated.