A NN-Based Control Method of Uncertain System with Large Time Delay
Aimed at the uncertain models in time-delay systems, a Smith prediction control scheme based on neural network was presented in this paper to overcome the drawback that traditional Smith predictor must depend on the mathematical model of the controlled pl
- PDF / 837,766 Bytes
- 7 Pages / 439.37 x 666.142 pts Page_size
- 3 Downloads / 220 Views
A NN-Based Control Method of Uncertain System with Large Time Delay Li-Xin Wei, Liang Cheng and Ying Li Abstract Aimed at the uncertain models in time-delay systems, a Smith prediction control scheme based on neural network was presented in this paper to overcome the drawback that traditional Smith predictor must depend on the mathematical model of the controlled plant. The scheme can make up for the poor robustness caused by the traditional Smith predictor when the parameters are changed. The simulation results show that this method could be applied in the case of unknown mathematical model of the system, and it had good adaptability when the system parameters were altered. Keywords Large time delay • Smith predictor • Neural networks • Uncertain system
18.1 Introduction To settle the time lag existed in the industrial production system, Smith predictor is commonly used. It can eliminate the time lag theoretically and improve the stability and response characteristic of the system. Considering the shortcomings that Smith predictor not only relies on the mathematical model of the system, but it is very sensitive to the errors of the model parameters. In Refs. [1–4], the structure of Smith predictor has been improved, it raised the anti-interference performance of the control scheme; Literature [5] combined internal model control and fuzzy control on the basis of Smith predictor, which increased the robustness of Smith predictor; Literature [6, 7] used the way of putting the fuzzy control and adaptive control combined to add on Smith predictor to reduce the dependence on mathematical models; Ref. [8] combined the controller of Smith predictor with the neural network partly. When the object model is completely unknown in the actual production, all of the methods are applied inefficiently.
L.-X. Wei · L. Cheng (*) · Y. Li Institute of Electrical Engineering, Yanshan University, Qinhuangdao 066004 , Hebei, China e-mail: [email protected]
W. Du (ed.), Informatics and Management Science I, Lecture Notes in Electrical Engineering 204, DOI: 10.1007/978-1-4471-4802-9_18, © Springer-Verlag London 2013
133
134
L.-X. Wei et al.
In this paper, a neural network-Smith (NN-Smith) predictive control scheme was proposed, which could make the controlled plant and its mathematical model matched real-timely through using the neural network to complete the identification and the tracking to the system parameters. As the simulation results showed, NN-Smith predictor is better than the common one both in the application and control effect.
18.2 Smith Predictor Control Smith predictive compensator control scheme can be described as follows: the dynamic characteristics of the process are pre-estimated under the basic disturbance, and then in order to make the hysteretic regulated variable reflected on the regulator in advance and take action at the same time, compensation control is put to use by Smith predictor, which can reduce the overshoot and accelerate the adjustment process. Suppose G 0 (s)e−τ s is the con
Data Loading...
