发酵过程混合神经网络模型及其仿真

A Hybrid Neural Network Model and Simulation for Fermentation Processes

提出了一种新型结构的发酵过程混合神经网络模型。该模型由非线性神经网络和线性神经网络两部分组成。由于非线性神经网络采用结构具有线性形式的Flat网络,两个网络能够合并为同一表达式,并具有线性形式,可采用线性最小二乘法求解网络权值。与串联结构及串并联结构混合神经网络模型相比,该模型训练方式简单,并可方便地使用在线辨识算法。

A new hybrid neural network model for fermentation processes is proposed, which combines the nonlinear network and the linear network. The nonlinear neural network is the Flat network that can be formulated as linear form. The nonlinear network is merged into a linear formula with the linear network. Thus, this formulation makes it easy to update the weight values of the network using a linear least-square method. Compared to existing models containing serial and serial-parallel hybrid neural network approaches in which more costly training is needed, the proposed model is very attractive if accuracy and easy training are critical issues.