基于BP神经网络和遗传算法优化番茄红素发酵培养基

Medium Optimization for the Production of Lycopene Based on BP Neural Network and Genetic Algorithms

DOI：10.3969/j.issn.1673-1689.2019.02.016

中文关键词: 神经网络 遗传算法 番茄红素 三孢布拉霉

英文关键词: neural network,genetic algorithms,lycopene,Blakeslea trispora

基金项目:

摘要点击次数: 282

全文下载次数: 287

中文摘要:

为了进一步提高三孢布拉霉高产突变株的番茄红素产量，本研究基于BP神经网络和遗传算法对发酵培养基的组成进行了优化。比较不同碳源、氮源、植物油对番茄红素产量和生物量的影响，确定最佳碳源、氮源和植物油。利用49组样本数据，建立以玉米粉、玉米浆、大豆油、磷酸二氢钾、硫酸镁为输入变量，番茄红素体积产量为输出变量的BP神经网络，并以建好的BP神经网络模型为适应度函数，利用遗传算法进行寻优。经过优化，得出番茄红素最大预测产量为1.27 g/L，经验证，与实际产量误差在5%以内，较优化前提高了31.6%。此时，玉米粉、玉米浆干粉、大豆油、磷酸二氢钾、硫酸镁的含量分别为41.2、8.93、26.5、1.39、0.46 g/L。因此，BP神经网络结合遗传算法是番茄红素发酵培养基优化的有力工具，番茄红素产量显著提高。

英文摘要:

The present study aims at using error Back Propagation（BP） neural network and genetic algorithms to improve lycopene production from semisynthetic medium by Blakeslea trispora. The effects of different kinds of carbon source，nitrogen source and vegetable oil on lycopene concentration and biomass are analyzed to confirm the component of medium. The establishment of BP neural network model is based on 49 group of samples data. In this model，Corn flour，corn steep liquor，soybean oil，monopotassium phosphate，magnesium sulfate are set to inputs，and lycopene volumetric production as output. Then，the genetic algorithms is applied to search the optimal value using BP network model as fitness function. After optimization，the maximum predicted value of lycopene production is 1.27 g/L. And the error of predicted value and actual value is less than 5% by experimental verification. Lycopene production in optimized medium is 31.6% higher than that in initial media. The optimized medium containes 41.2 g/L corn starch，8.93 g/L corn steep liquor，26.5 g/L soybean oil，1.39 g/L KH2PO4，0.46 g/L MgSO4. The combination of BP neural network with genetic algorithms is a power tool to obtain optimized lycopene fermentation medium.

查看全文 查看/发表评论 下载PDF阅读器