Abstract Within this work a nonlinear model predictive controller (MPC) was implemented in a Penicillium chrysogenum fed-batch process and compared to a PI(D) and an open loop feedback control scheme, referenced as model based control (MBC). The controllers were used to maintain predefined set-points of biomass specific glucose uptake rates, product precursor and nitrogen concentrations by manipulating the glucose, precursor and nitrogen feeds. As the critical component concentrations are not available for direct measurement a particle filter including measured oxygen uptake rate (OUR) and carbon evolution rate (CER) was deployed to estimate biomass, nitrogen and product precursor concentrations. State estimation and predictive control actions were based on a kinetic model which was retrieved from literature and adapted to the examined process and control tasks by simplifying the description of the hyphal compartmentalization and adding nitrogen as well as the measurable OUR and CER. Besides simulations, verification experiments of the developed control schemes were executed. Although the kinetic model used for state estimation and prediction does not reflect the overall biological complexity it could be successfully used to estimate and control the glucose uptake and the unmeasured component concentrations. During experimental verification, nonlinear process dynamics caused unstable PI(D) behavior. In comparison to PI(D) and MBC, the MPC efficiently avoided formation of by-products, which resulted in efficient substrate utilization and an overall product gain of 14%.

中文翻译：

---

黄青霉分批补料过程中模型预测、PID和模型反演控制的实验验证与比较

摘要 在这项工作中，非线性模型预测控制器 (MPC) 在 Penicillium chrysogenum 分批补料过程中实施，并与 PI(D) 和开环反馈控制方案进行比较，称为基于模型的控制 (MBC)。控制器用于通过操纵葡萄糖、前体和氮进料来维持生物质特定葡萄糖摄取率、产物前体和氮浓度的预定义设定点。由于关键组分浓度不可用于直接测量，因此部署了包括测量的吸氧率 (OUR) 和碳析出率 (CER) 的粒子过滤器来估计生物量、氮和产物前体浓度。状态估计和预测控制动作基于从文献中检索到的动力学模型，并通过简化菌丝分隔的描述和添加氮以及可测量的 OUR 和 CER 来适应检查的过程和控制任务。除了模拟之外，还对所开发的控制方案进行了验证实验。尽管用于状态估计和预测的动力学模型不能反映整体生物复杂性，但它可以成功地用于估计和控制葡萄糖摄取和未测量的组分浓度。在实验验证期间，非线性过程动力学导致不稳定的 PI(D) 行为。与 PI(D) 和 MBC 相比，MPC 有效地避免了副产物的形成，