人工神经网络用于乙醇发酵生产过程的建模探讨  

Application of artificial neural network in modeling ethanol fermentation process in factory

作　　者：柳元[1] 胡山鹰[1] 陈定江[1] 沈静珠[1] 杜风光[1] 

机构地区：[1]清华大学化学工程系生态工业研究中心,北京100084

出　　处：《计算机与应用化学》2009年第8期998-1002,共5页Computers and Applied Chemistry

基　　金：国家自然科学基金重点项目(20436040)

摘　　要：基于发酵生产的特点及建模要求,以某企业燃料乙醇生产过程为研究对象,利用工业生产中的参数及数据,建立了以乙醇发酵效果为目标的BP神经网络模型,以静态模型反映复杂的动态问题。探讨了乙醇发酵生产模型的误差产生原因,并提出改进方案,根据已有经验将相关参数进行了合理组合,调整神经网络模型的输入输出参数结构,以提高仿真模拟效果。通过多次模型改进,使模拟的平均相对误差从10%提高至5.4%,为进一步研究发酵生产建模提供了思路。With analyzing characteristics of industrial fermentation process, the artificial neural networks in static modeling of ethanol industrial fermentation have been attempted, The parameters in ethanol fermentation process were firstly investigated, with the input and output parameters of the models selected. Based on the historical production data from a bio-ethanol factory, a BP neural network model was trained, and improvements for another two models, with some reasonable combinations of selected parameters, were discussed to achieve satisfactory simulation results. The last result of the mean relative error has been raised from 10% to 5.4%, which is helpful for the further modeling research of industrial fermentation.

关 键 词：发酵建模 神经网络 参数分析 乙醇产出模型 

分 类 号：TQ015.9[化学工程] TP391.9[自动化与计算机技术—计算机应用技术]