基于神经网络的生化过程预估优化控制

综观历史与现状,生化过程数学模型的建立是很困难的。不多的一些数学模型也往往由于精度低,应用范围窄而无法在实际中应用。这是由于生化过程的机理非常复杂,具有高度非线性和时变特性。并且不同于一般物理过程的是生化过程是个物理上不可逆的过程。近几年来,人工神经网络(ANN)得到了迅速发展,并被广泛应用到各个领域。同样,ANN也为生化过程控制提供了一种新方法。本文以工业生产中发酵过程的补料控制为例,叙述了ANN如何用干生化过程预估和优化控制。相应地,对结果进行了分析讨论。

Prediction and Optimal Control for Bioprocess Based on ANNs

Bioprocess is highly complex process with high variability and little knowledge about its mechanisms. It is very difficult to get methematical models that are accurate enough for state estimation, modelling and optimization. In this paper, the approach for prediction of a fed- batch fermentation based on artificial neural networks is introduced in first. Then, the method of feedback optimization of this process is presented. Finally, an example of industrial application is introduced, and the result is discussed.