发酵过程的建模与优化方法研究

Modeling and optimization method research of the fermentation process

对于发酵这样一个复杂的非线性动态过程,由于在线传感器的缺乏,使得过程中的一些重要状态变量难以在线测量,从而给发酵过程的优化控制带来了极大困难。为此,提出了一种新型的动态网络—递归补偿模糊神经网络方法,实现对发酵过程的建模和状态估计,结果表明该网络能够较为准确地拟合过程的动态特性。进一步采用改进的蚁群算法来对发酵过程的控制变量进行优化,使发酵的产物产量得到提高。该方法应用于多粘菌素的发酵生产过程中,实现了状态变量的在线预估与控制变量的在线优化。

Because of the complexity of the process and the lack of biosensor for the fermentation process, some important variables can＇t be measured on line. The optimization control of the process is difficult. So a new dynamic neural network- recurrent fuzzy neural network was proposed. It was applied to model and predict the variables. The result has shown that the neural network models can approximate the dynamic characteristic of the process perfectly. Based on the neural network models, the control variables are optimized with improved ant algorithm. And the yield of the product can be increased by optimization. It can realize the online prediction of the variables and the optimization of the control variables in mycetozoan fed-batch process.