Study of advanced control of ethanol production through continuous fermentation

详细信息    查看全文

• 作者：AbdelHamid Ajbar ^{amkamal@ksu.edu.sa} ; Emad Ali ; ^{aajbar@ksu.edu.sa}
• 关键词：Ethanol fermentation ; Reactor model ; Nonlinear model predictive control ; Proportion ; Proportional integral control ; Maximum productivity
• 刊名：Journal of King Saud University - Engineering Sciences
• 出版年：2017
• 出版时间：January 2017
• 年：2017
• 卷：29
• 期：1
• 页码：1-11
• 全文大小：874 K
• 卷排序：29

文摘

This paper investigates the control of an experimentally validated model of production of bioethanol. The analysis of the open loop system revealed that the maximum productivity occurred at a periodic point. A robust control was needed to avoid instabilities that may occur when disturbances are injected into the process that may drive it toward or through the unstable points. A nonlinear model predictive controller (NLMPC) was used to control the process. Simulation tests were carried out using three controlled variables: the ethanol concentration, the productivity and the inverse of the productivity. In the third configuration, the controller was required to seek the maximum operating point through the optimization capability built in the NLMPC algorithm. Simulation tests presented overall satisfactory closed-loop performance for both nominal servo and regulatory control problems as well as in the presence of modeling errors. The third control configuration managed to steer the process toward the existing maximum productivity even when the process operation or its parameters changed. For comparison purposes, a standard PI controller was also designed for the same control objectives. The PI controller yielded satisfactory performance when the ethanol concentration was chosen as the controlled variable. When, on the other hand, the productivity was chosen as the controlled output, the PI controller did not work properly and needed to be adjusted using gain scheduling. In all cases, it was observed that the closed-loop response suffered from slow dynamics, and any attempt to speed up the feedback response via tuning may result in an unstable behavior.