基于铝电解过程神经网络建模的研究

Modeling Research of Neural Network Based on Aluminum Electrolysis

目的利用神经网络铝对电解过程进行辨识建模,以解决采用常规方法难以建立模型的问题.方法分别利用递归神经网络(Elman神经网络)和延时神经网络(加入延时单元的BP神经网络)对铝电解过程进行辨识建模,并将二者的辨识结果进行了比较.结果递归Elman神经网络能更好地跟踪铝电解生产过程,并且网络结构简单误差小.结论笔者提出的递归El-mall神经网络建模方案更适合于对铝电解过程进行辨识建模.

Using neural network can establish the model of aluminum electrolysis. That can solve the problem of aluminum electrolysis process modeling by using regular method. This paper presents aluminum electrolysis process modeling method by using partial recurrent neural network（Elman neural network） and time delay neural network（BP neural network with time delay unit） respectively . The identification results which obtained by Elman neural network and TDBP neuralnetwork are compared. The results show the Elman neural network can trace the aluminum electrolysis process better than the time delay neural network. And the Elman neural network has simple strcture and little error of identification.