[4Rin1] Interactive 2

[4Rin1-87] Inferential Control of Distillation Column by Successive Update of Softsensor and Control Algorithm

〇Masanori Oshima¹, Sanghong Kim¹, Ken-ichiro Sotowa¹ (1.Kyoto University)

Keywords:Softsensor, Inferential control, Chemical process

Inferential control is a control method where the value of the controlled variable is predicted by a softsensor, a statistical model, and transmitted to a controller. A softsensor is usually built with past operating data obtained from an existing plant. The prediction accuracy of the softsensor will deteriorate when the condition where the model construction data was obtained is different from that where prediction is required. The operating condition of the inferential control system is different from that of the existing plant because the controlled variable of the inferential control system is different from that of the existing control system. When the operating condition change occurs, the control performance of the inferential control system will improve, but we need to pay attention to the deterioration of the prediction accuracy. In the present work, we aimed to propose a novel method to design the inferential control system, considering the trade-off between the control performance improvement and the prevention of the prediction accuracy deterioration. The performance indexes of control performance and prediction accuracy in quality inferential control of a distillation column were improved by 66.6% and 98.2% respectively compared to a conventional method.