Identification And Control For Time-delay Systems

The time-delay system is very common in industrial processes and its control performance has a great impact on product quality. The existence of time-delay makes the controller not respond the disturbance in time, resulting in significant overshoot and big adjustment time and worsening the system stability. In general, it is difficult to control time-delay systems. In this paper, the frequency domain method is used to identify the rough time-delay system model and the improved genetic algorithm is utilized to get the accurate model.The Smith predictor is used to compensate time-delay. It can be combined with the GPC algorithm to help the time-delay system obtain better tracking performance and robustness.First, common approaches to get the system frequency characteristics are introduced, among which the spectral analysis method is described in detail.The system transfer function is fitted to the frequency characteristics and the model structure and initial parameters are prepared for further identification.Second, the basic principles and common modified measures of Smith predictive control and genetic algorithm are introduced.A modified genetic algorithm is presented. It adopts the appropriate coding and genetic manipulation strategies and dynamically adjusts the crossover probability and mutation probability based on fuzzy rules in the algorithm process. This method is used to identify the model parameters in the Smith predictor and simulation proves that good control performance can be achieved.Third, the GPC algorithm is modified by adding the Smith predictor and the algorithm theory and parameter selection principles are described. Simulation results show that this algorithm not only can get excellent dynamic control performance but also has good real-time and robustness properties and is not sensitive to the model mismatch.Finally, an interface program is designed using MATLAB and all kinds of identification and control algorithms are integrated, which facilitates the simulation of the algorithms and field data processing.