摘要
为复杂的发酵过程建立软测量模型要求模型最好能够给出预测值的置信区间,以便技术人员对发酵过程的真实状况和模型的可靠性进行评估。贝叶斯极限学习机能够在实现预测的同时一并给出预测值的置信区间,因此将其用于发酵过程的软测量建模。然而,实际发酵过程中的输入数据往往带有噪声,贝叶斯极限学习机仅能处理输出含噪声的情况。针对这个问题,提出了输入不确定贝叶斯极限学习机。在原有的贝叶斯推理过程中引入输入不确定性,得到了综合考虑输入输出噪声的模型参数和预测置信区间。最后利用青霉素发酵过程进行仿真验证,建立了产物质量浓度的软测量模型,结果表明该方法预测精度高,得到的预测置信区间包含了所有真实值。
Soft sensor for fermentation process is better to get the confidence interval of prediction value. So that technical staff can make reasonable evaluation on the real situation of fermentation process. Bayesian extreme learning machine is applied to the soft sensor of fermentation process because it can obtain prediction value and confidence interval at the same time. However, noisy input is not considered by the modeling meth-od based Bayesian extreme learning machine, this is not consistent with real condition. To address this prob-lem, Bayesian extreme learning machine with input uncertainty is proposed. It is a Bayesian extreme learning machine framework which allows for input noise and get the model parameters which are overall consideration of input and output noise as well as confidence interval of prediction. Applying the proposed method to model the penicillin fermentation process, the simulation results show its good prediction accuracy and all of real val-ues are included into the confidence interval.
出处
《自动化与仪器仪表》
2014年第5期122-126,129,共6页
Automation & Instrumentation
关键词
发酵过程
软测量
极限学习机
贝叶斯模型
输入不确定
置信区间
Fermentation process
Soft sensor
Extreme learning machine
Bayesian Model
Input uncer-tainty
Confidence interval