基于神经网络与遗传算法优化γ-氨基丁酸的发酵条件

Fermentation conditions optimization for γ-aminobutyric acid production based on neural network and genetic algorithms

本文运用BP(back-propagatfon)神经网络优化红曲霉ZL307产γ-氨基丁酸的固态发酵工艺条件,建立了发酵条件与γ-氨基丁酸产量的关系模型,采用遗传算法对此模型进行全局寻优.神经网络结构为4-11-1的模型能较为精确地拟合输入的样本数据,测试样本的输出值与试验结果的相关系数为0.989.遗传算法优化出的最佳工艺参数为温度31.7℃、初始pH 4.6、初始含水量69.8%,接种量13.2%.在优化条件下,γ-氨基丁酸产量为0.518mg/g,含量比优化前提高了19.6%.

The solid fermentation conditions tor γ-ammobutyrm acla prouuction by ,wu optimized. Back-propagation neural network was established to model the relationships between the 7- aminobutyric acid yield and fermentation conditions. Genetic algorithm（GA）was applied for integrated optimization of the model. The trained network with the structure of 4-11-1 had a high generalization, the correlation coefficient between simulating outputs and the test data was 0. 989. The obtained optimal conditions were：temperature 31.7 ~C ,initial p H 4.6, initial water content 69.8 %, and inoculum volume 13.2%. Under above conditions, the yield of γ-aminobutyric acid reached 0. 518 mg/g, 19.6% higher than before.