About Us
Grad Studies
Theme Topic
Real Time Optimization

In order to excel in the present environment of increased market competition and rising costs, industries need to extract the maximum attainable economic performance from their operations subject to product quality, safety, environmental and other prevailing constraints. Real-time optimization (RTO) is a mechanism toward achieving this goal through continual adjustment of the plant operating point to coincide with the economic optimum. The process systems group at McMaster University has been at the forefront of research in RTO systems. (Marlin and Hrymak, 1997; Yip and Marlin, 2004). While significant advances have been made in this area, several challenges remain:

  • Most RTO systems operate in conjunction with model predictive control. However, the interaction between these two control levels has not been fully explored. In particular, industrial MPC systems are generally implemented in conjunction with a linear programming (LP) or quadratic programming (QP) steady-state optimizer. Despite the apparent success of two-level LP-MPC systems, instances of poor performance have been reported. In a recent study, we have shown that the LP can have the effect of amplifying measurement and/or process noise (Nikandrov and Swartz, 2009). Further planned studies include analysis of the effect of the control structure on the performance of LP-MPC cascade systems, and development of a framework for the design of LP-MPC systems.

  • Most real-time optimization strategies use steady-state models to determine optimal set points which are transmitted to an underlying regulatory control layer. However, many processes operate under conditions that are intrinsically transient in nature, such as batch operations, grade transitions in continuous operations, start-up and shut-down. Occurrences of intentionally transient operation are moreover increasing with increasing trends toward market-driven operation, where plants need to respond to rapidly changing customer demands. This suggests the consideration of dynamic effects at the RTO level.

  • RTO systems typically utilize high-fidelity process models. However, this may pose a computational bottleneck for large systems, particularly if dynamic models are considered. This suggests an investigation into the use of reduced-order surrogate models, with appropriate model adaptation schemes.

Dr. Chris L. E. Swartz
Professor and Director, MACC
Dr. Thomas E. Marlin
Professor Emeritus
Mahdi Eskandari
MaSC recipient
Hao Li
MaSC recipient
Dynamic real-time optimization with closed-loop prediction
AIChE J (2017)  -  [ Publisher Version ]
Globally optimal dynamic real time optimization without model mismatch between optimization and control layer
Xiaoqiang Wang, Mahalec, V., Feng Qian
Comp. Chem. Eng, 104 64-75 (2017)  -  [ Publisher Version ]
Approximation of closed-loop prediction for dynamic real-time optimization calculations
Computers & Chemical Engineering, 103 23-38 (2017)  -  [ Publisher Version ]
Globally optimal nonlinear model predictive control based on multi-parametric disaggregation
Xiaoqiang Wang, Mahalec, V., Feng Qian
Journal of Process Control, 52 1-13 (2017)  -  [ Publisher Version ]
A Multi-rate Moving Horizon Estimation Framework for Electric Arc Furnace Operation
IFAC-PapersOnLine, 49 (7) 11751180 (2016)  -  [ Publisher Version ]