基于通用学习网络的控制器设计

Controller Design Based on the Universal Learning Network

神经网络已经广泛地应用于控制领域,但是现有的神经网络一般不包含延迟时间,或延迟时间不能任意设置。通用学 习网络(UniversalLearningNetwork)是一种新型可设定神经网络,该网络所有节点互连,任意两节点之间可以有多重连接, 且连接上有任意的延迟时间。利用通用学习网络对典型的非线性、大滞后系统——pH中和过程进行辨识,并作为预估器应用 于控制系统。仿真研究证明该网络能有效地应用于大滞后系统。

Artificial neural network has been widely used in control field, but most networks at present do not contain time-delay, or the time-delay is constant and is not arbitrary. A new type of neural network-universal learning network (ULN) is introduced, the nodes of the network is connected each other, and between any two nodes, there are multi-branch, and also the time-delay can be set arbitrarily on each branch. ULN is adopted to identify and control the typical nonlinear and long time-delay system-pH neutralization process and apply as predictor in control system. It is proved that the ULN can be effectively used in nonlinear and long time-delay system.