金霉素发酵过程优化调度策略研究  

Study on Optimal Scheduling Strategy of CTC Fermentation

作　　者：杨建文[1] 陈祥光[1] 金怀平 吴磊[1] 

机构地区：[1]北京理工大学化工与环境学院,北京100081

出　　处：《北京理工大学学报》2014年第8期876-880,共5页Transactions of Beijing Institute of Technology

基　　金：博士后科学基金资助项目(20100480208)

摘　　要：针对金霉素发酵过程影响优化控制的难测参数,提出了一种基于数据场聚类、模糊神经网络和滚动学习预报的优化调度策略.提取输入变量数据场聚类特征值作为模糊神经网络模型结构参数的初始值,消除人为参数选择的随机性误差,并在预测模型中加入离线数据模型修正算法.因此,提出的优化调度策略提高了对金霉素发酵过程难测变量的预测精度,增强了预报模型的鲁棒性.现场运行结果表明,提出的方案将企业金霉素生产的综合效益提高了9.12%,具有很好的应用价值.Considering these online unmeasurable parameters affecting the optimized control, an optimal scheduling strategy was proposed on the basis of data field clustering, fuzzy neural networks and rolling-learning prediction. With the eigenvalues of data field clustering information for input variables initializing model parameters of the fuzzy neural networks, this solution eliminated the stochastic error of artificial parameter selection. In addition, the algorithm of correcting the model making use of offline data was added into the prediction model. Therefore, the proposed strategy improved the estimation accuracy of these unmeasurable parameters of CTC fed-batch process and enhanced the robustness of prediction models. The experimental results show that the scheme increases the production efficiency of the CTC fermentation and takes on a promising practical application.

关 键 词：金霉素发酵 数据场聚类 效益函数 优化调度策略 滚动学习预报 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]