【作者】 褚丽丽；

【导师】 李春茂；

【作者基本信息】 西南交通大学 ， 电力电子与电力传动， 2008， 硕士

【摘要】 工业生产过程中,时滞对象普遍存在,同时也是较难控制的,尤其是大时滞对象的控制一直都是一个难题。而很多温度控制系统都是属于大时滞系统,常见的智能温度控制器虽然在温度控制的实际应用中表现了比较理想的控制效果,但它仍然属于将参数整定与系统控制分开处理的离线整定方法,如果工况发生变化就必须重新调整参数。针对这一问题,为了实现时滞系统参数自整定的控制,本文将神经网路控制、模糊控制和PID控制结合起来,设计了基于神经网路的模糊自适应PID控制器。首先,本论文分析了时滞系统的特点,讨论了几种时滞系统较为成熟的常规控制算法:微分先行控制算法、史密斯预估控制算法、大林控制算法,并深入研究了它们的控制性能;并且通过仿真对这三种控制方法在温控系统中的控制性能进行了比较。其次,在分析PID参数自整定传统方法的基础上,设计了一种改进方法,并设计了相应的控制器。该控制器综合了模糊控制、神经网络控制和PID控制各自的长处,既具备了模糊控制简单有效的控制作用以及较强的逻辑推理功能,也具备了神经网络的自适应、自学习的能力,同时也具备了传统PID控制的广泛适应性。该方法不需要离线整定参数,实现了在线自整定参数。仿真实验表明了该控制器对模型和环境都具有较好的适应能力和较强的鲁棒性。最后将基于神经网路的模糊自适应PID控制器应用于贝加莱PID温控装置,能够出色地实现参数的在线自整定。理论分析、系统仿真、实验结果都证实了这种控制策略能有效地减少系统超调量,并减少了调节时间,提高了系统的实时性和控制精度。更多还原

【Abstract】 Time-delay systems are very widely used in industry and are hard to control, especially when the plant with a long delay. However, a lot of temperature control systems belong to time-delay systems. Although the classical intelligent temperature controller in the practical application of temperature control shows ideal control effect, it still adopts offline tuning methods to solve parameter setting and to control system separately. If the situation changes the parameter must be readjusted. Aimed at the problem, by combining of neural network control, fuzzy control, and PID control, a fuzzy adaptive PID controller based on neural network is designed to adjust parameters online to achieve high performance.Firstly, this thesis analyzes the characteristics of time-delay system, discusses some general control schemes for time-delay, for example, differential forward control algorithm, Smith predictor control algorithm, Dahlin control algorithm. And it analyzes their control performance. And the result of simulation shows the difference of their performance in temperature control systems.Secondly, the thesis describes the traditional principle of PID parameter control, and designes a improved scheme and its relevant controller. This controller comprehensively combines the advantage of the neural network control, fuzzy control and PID control. It possesses not only the simplicity control and strong logical inference of the fuzzy control, but also the learning and adaptive functions by using the neural network. Furthermore, it is as widely adaptive as PID controller. It realizes the PID parameters’ tuning online and don’t need offline tuning. It has proven that this controller has adaptability and robustness to the environment.Finally, the thesis adopts the fuzzy adaptive PID controller based on neural network to the B&R temperature control system. That could finish the PID parameters’ tuning online and obtain perfect control effect. Theoretical analysis, simulation and experiment results have proven that the strategy could decrease the overshoot and shorten the tuning time, improve the system’s real-time performance and precision.更多还原