单神经元PD多变量解耦控制研究18.doc

单神经元PID多变量解耦控制研究 摘 要 对于具有非线性、大迟滞、强耦合特点的多变量系统，研究人员很难找到理想方法解决控制中的诸多问题。对于多变量系统之间的耦合，有些可以采取被调量和调节量之间的适当匹配，和重新整定调节器的方法加以克服。PID控制方法是经典控制算法中的典型代表，并在多种控制场合取得了很好的效果，但随着生产工艺的日益复杂和人们对工业过程总体性能要求的不断提高，传统的PID控制方法往往难以满足闭环优化控制的要求。基于知识且不依赖于模型的智能控制为解决这类问题提供了新的思路，成为目前提高过程控制质量的重要途经。而神经网络作为现代信息处理技术的一种，正在很多应用中显示它的优越性，它在自动控制领域的应用成果---神经网络控制也成为令人瞩目的发展方向。单神经元作为构成神经网络的基本单位，具有很强的信息综合、学习记忆和自学习、自适应能力，可以处理那些难以用模型和规则描述的过程，而且结构简单易于计算。若将这两者结合，则可以在一定程度上解决传统PID调节器不易在线实时整定参数、难于对一些复杂过程和参数慢时变系统进行有效控制的不足。正是利用它们的优点做成单神经元自适应PID控制器对多变量系统进行解耦控制会起到一个很好的控制效果。 关键字：解耦控制系统；多变量解耦；单神经元自适应PID The Research Of Single Neuron PID Multivariable Decoupling Control Abstract For the nonlinear, heavy delay, the strong coupling characteristics of multivariable systems, Researchers are difficult to find an effective control strategy. For multivariable systems, the coupling, and some can be taken to adjust capacity and transfer the appropriate amount of matching, and re-tuning regulator approaches to overcome. PID control method is one of the traditional control methods and gets good effects under many application situations. But with the increase in complexity of manufacture technology and demands of industrial process performance, the conventional PID control can not meet the requirement of closed loop optimized control, Intelligent control independent of model of a plant and based on knowledge offers a new idea for improving the process control quality, of which neural network as one of modern information process technologies, has some advantages in many applications. Neural network control became a regarded research direction. Single neuron as a neural network the basic unit, has the very strong ability in information synthesis, study memory, self-study, and adaptation, so, it can deal with some processes that are difficult to describe with the model or rule, structure is simple and calculation is very easy. * If they combination, they can to some extent solve the traditional PID controller difficult onl