论文部分内容阅读
本文研究了一般化学习网络(Universal Learning Network)在多变量连续釜式反应器(CSTR)系统的建模应用.一般化学习网络具有节点之间有多重分支、任意2个节点互连且节点之间可具有任意的时间延迟的特点,因此能够应用在高度非线性复杂系统的辨识中.分别用一般化学习网络和常规的递归神经网络对多变量连续釜式反应器(CSTR)进行系统辨识比较,仿真结果验证了一般化学习网络结构比递归神经网络Elman的辨识精度更高,且网络结构更简洁紧凑的特点.
This paper studies the application of Universal Learning Network in the modeling of multivariable continuous tank reactor (CSTR) system.Generalized learning network has multiple branches between nodes, any two nodes are interconnected and nodes So it can be used in the identification of highly nonlinear and complex systems.Using generalized learning networks and conventional recurrent neural networks to systematically identify and compare multivariable continuous tank reactors (CSTRs) The simulation results verify that the generalized learning network structure is more accurate than the recursive neural network Elman and the network structure is more compact and compact.