基于神经网络自学习的PID控制算法研究学士学位论文.doc

学士学位论文 PAGE \* MERGEFORMAT PAGE \* MERGEFORMAT V 摘 要 PID（比例-积分-微分）控制器作为最早实用化的控制器已有50年以上的年历史，现在仍然是应用最广泛、最普遍的工业控制器。PID控制器简单易懂，使用中不需精确的物理系统模型等先决条件，因而成为最受欢迎的、应用最为普遍的控制器。PID控制器最早发展起来的原因，是由于其算法简单、鲁棒性好和可靠性高，被广泛应用于过程控制和运动控制中，特别适用于可建立精确数学模型的确定性控制系统。但是常规的PID控制器参数往往整定欠佳、性能不良，对运行环境的适应性很差。而神经网络具有很强的非线性映射能力、自学习的能力、联想记忆的能力、并具有可行批量信息处理方式及良好的容错性能。 本次本课题设计的目的就是为了建立一种单神经网络的PID控制器，使得人工神经网络与传统PID控制相结合互相补充，共同提高控制质量、以及控制效率并利用Matlab软件进行仿真。 关键词：PID控制；神经网络PID；RBF算法；MATLAB仿真 ABSTRACT The PID ( PID ) controller as the first practical controller has 50 years of history, is still the most widely used industrial controller. The PID controller is simple and easy to understand, without the use of accurate system models prerequisites, and thus become the most widely used controller. The PID control room is the earliest developed one of the control strategy, because of its simple algorithm, good robustness and high reliability, is widely used in process control and movement control, especially can be applied to establish the precise mathematical model of uncertainty control system. But the conventional PID controller parameters often setting bad, poor performance, the operating condition adaptability is poor. While the neural network has very strong nonlinear mapping ability, self-learning ability, the capacity of associative memory, parallel information processing and fine fault-tolerant performance. This topic design proposed is to establish a single PID neural network controller, the artificial neural network and traditional PID control are combined to complement each other, work together to improve the control quality, and the use of Matlab software simulation. Key words: PID control;neural network PID ;RBFalgorithm ; MATLAB simulation 目 录 TOC \o 1-3 \h \u HYPERLINK \l _Toc13036 1 绪论 HYPERLINK \l _Toc26938 1.1 课题研究背景 PAGEREF _Toc26938 1 HYPERLINK \l _Toc3682 1.2 课题研究意义 PAGEREF _Toc3682 2 HYPERLI