基于神经网络和遗传算法培养基优化的发酵经济学

The Economics of Fermentation for Medium Optimizing by Neural Networks Coupling Genetic Algorithms

利用目前较为先进的神经网络(ANN)和遗传算法结合(GA)的优化方法,通过以6种培养基组成为输入,亚硝酸盐氧化菌的活性与培养基成本之比的"性价比"为输出,构建结构为6-8-1BP神经网络的非线性的非结构模型,并以该模型为遗传算法的目标函数,进行遗传算法的全局寻优,优化得到具有最高"性价比"的培养基组成。结果表明,基于神经网络和遗传算法能较好地对培养基进行优化,并实现对发酵经济学的初步研究。通过优化,得到亚硝酸盐氧化菌的最佳培养基组成为:NaNO22.390 g/L,KH2PO41.355 g/L,MgSO40.019 g/L,NaCl 0.031 g/L,NaHCO34.373 g/L,FeSO40.005 g/L,其最优"性价比"为8.705,比初始的6.835提高了27.36%。

The aim of study the economics of fermentation is to control the cost of products and increase the rate of product market share by reducing the cost and raising the performance--price ratio. In this study, the economics of fermentation medium was evaluated by applying of neural networks coupling with genetic algorithms. An optimal medium composition of nitrite oxidizing bacteria： NaNO2 2. 390 g/L, KH2PO4 1. 355 g/L, MgSO4 0.019 g/L, NaCl 0.031 g/L, NaHCO3 4. 373 g/L, FeSO4 0. 005 g/L, and the optimal performance price ratio were achieved at 8. 705, higher 27.36% than that of the corresponding value of the initial （6. 835） in a limited time and more less of experiments.