10th IFAC Symposium

on Advanced Control of Chemical Processes

July 25 - 27, 2018 Shenyang, China


Real-Time Optimization with Modifier Adaptation

Please click here to download workshop slides.


Sebastian Engell, Process Dynamics and Operations Group, Technische Universit├Ąt Dortmund, Dortmund, Germany

Dominique Bonvin, EPFL, Lausanne, Switzerland


Sebastian Engell, Afaq Ahmad, Reinaldo Hernandez, TU Dortmund

Dominique Bonvin, Martand Singhal, EPFL Lausanne



Workshop Objectives

The goal of the workshop is to familiarize the attendees with the modifier-adaptation strategy and the recent developments that make the approach very attractive for industrial applications. This workshop will be of interest to process control engineers, students and young investigators who want to consolidate their knowledge in this area.


Companies in the process industries face intense pressure to improve production efficiency, while maintaining product quality and process safety. One means to achieve this goal is real-time optimization (in short RTO). RTO aims at iteratively improving the operation of continuous (essentially stationary) processes as well as of cyclic processes and repeated batch processes. As a model-based optimization approach, the success of RTO depends strongly on the quality of the models used for optimization. The effort required for building and maintaining models is the bottleneck in the deployment of RTO solutions, and even when sophisticated models are used, they never exactly represent the real process.

A promising RTO strategy is modifier adaptation (MA). The effect of plant-model mismatch on the optimization is compensated by zeroth- and first-order corrections to the cost and constraint functions of the optimization problem. The so-called modifiers are calculated from data obtained at previously visited operating points. Simulation and experimental studies have shown that this strategy is effective to handle plant-model mismatch and drive the process to optimally.

Taught jointly by two of the most active research groups in this area, this workshop will present a comprehensive overview of MA, including some historical background, theoretical foundations (optimality and convergence), practical aspects (gradient estimation), as well as some of the latest developments (Dual MA, Directional MA, MAWQA, MA based on transient data) and applications.




Motivation: Optimizing operations in the process industry under model uncertainty

Sebastian Engell



Modifier Adaptation: Fundamentals

Dominique Bonvin and Martand Singhal

- Problem formulation
- KKT matching, optimality
- Gradient estimation and dual modifier adaptation
- Convergence, feasible-side convergence
- Trust-region methods




Implementation aspects

Modifier adaptation combined with quadratic approximation

Sebastian Engell

Speeding-up real-time optimization with modifier adaptation

Reinaldo Hernandez

Directionality aspects in modifier adaptation

Martand Singhal

Efficient parameter estimation

Afaq Ahmad


Coffee break



Real-world applications of modifier adaptation

Kite for energy harvesting

Dominique Bonvin

Fuel-cell system

Martand Singhal

Hydroformylation miniplant

Reinaldo Hernandez



Discussion: Challenges and new avenues