基于QPSO-ANN的发酵过程建模与优化  

Modeling and Optimization for Fermentation Process Based on QPSO-ANN

作　　者：杜润龙[1] 须文波[1] 

机构地区：[1]江南大学信息工程学院,江苏无锡214122

出　　处：《微电子学与计算机》2007年第11期99-102,共4页Microelectronics & Computer

基　　金：国家自然科学基金项目(60474030)

摘　　要：针对BP神经网络训练时间长、易陷入局部极小点问题,将量子微粒群算法QPSO与BP算法结合起来分两次训练神经网络,建立青霉素浓度预估模型。用青霉素发酵数据集对模型进行训练与检验。基于该模型,用QPSO算法对温度与pH控制轨线进行优化。实验表明,该发酵过程模型训练误差小、学习速度快、泛化能力强、预测精度高、可以实现多步预估。采用优化后的温度、pH控制轨线,青霉素浓度有所提高。Due to BP neural networks is defective in rapidity of convergence and apt to trap into local extreme value, this paper bring forward a kind of second training method of the neural network by combining BP algorithm with QPSO algorithm for estimation model of penicillin fermentation concentration. Training and testing the model with penicillin fermentation data-set. Based on this model, QPSO algorithm is applied to optimize trajectories of temperature and pH in each fermentation phase. The experiment results indicates that this model features small training error, high learning speed, well generalization ability, high estimation precision. The model is able to realize multi-step pre-estimate. Optimizing control trajectories of temperature and pH makes the fermentation concentration increase.

关 键 词：微粒群算法 量子 发酵过程 建模 优化 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]