摘要
青霉素发酵过程具有较强的非线性、时变性和不确定性,发酵过程中的基质浓度、青霉素菌体浓度、产物浓度等关键生物参数难以实时在线测量,而离线化验存在时滞大的问题,难以满足实时在线控制的要求。针对这一问题,提出了一种基于核主成分分析(KPCA)与支持向量机回归(SVR)的软测量建模方法。首先,利用KPCA提取软测量输入数据空间中的非线性主成分;然后,采用SVR算法建立了可准确预测青霉素发酵过程重要参数的软测量模型。试验结果表明,与传统建模方法相比,KPCA-SVR软测量模型的测量精度高、跟踪性能好、泛化能力强,能满足发酵过程中青霉素菌丝浓度的在线测量要求,是一种有效的软测量建模方法。
Penicillin fermentation process is one kind of highly nonlinear,timevarying and uncertain process,and its key biological parameters including substrate concentration,penicillin cell concentration,product concentration are difficult to be measured in realtime and online,while the offline assay has large time delay,it is difficult to meet the requirements of realtime and online control.Aiming at this problem,a soft sensing modeling method based on kernel principal component analysis （KPCA） and support vector regression （SVR） is put forward.By adopting KPCA,the nonlinear principal component in input data space of soft sensing is extracted; then the soft sensing model which can accurately predict the important parameters of penicillin fermentation process is established by using SVR algorithm.The experimental results show that compared with the traditional modeling methods,the KPCASVR soft sensing model features higher measuring accuracy,better tracking performance and stronger generalization ability.It can meet the requirements of online measurement of penicillin mycelium concentration in fermentation process,which is an effective soft sensing modeling method.
出处
《自动化仪表》
CAS
2018年第2期12-16,共5页
Process Automation Instrumentation
基金
中央高校基本科研业务费专项基金资助项目(15CX02103A)
中国石油大学胜利学院科技计划基金资助项目(ky2017006)
关键词
生物发酵
青霉素
核主成分分析
支持向量机回归
软测量
非线性模型
Biological fermentation
Penicillin
Kernel principal component analysis （ KPCA ）
Support vector regression（SVR）
Soft sensing
Nonlinear model
