Keynote Speakers

The Keynote Speakers in the alphabetical order:

Cesar De Prada (University of Valladolid, Spain)

Cesar De Prada

Title: Modifier adaptation methods for RTO

Abstract:

Several approaches have been developed to deal with the mismatch between process and model in Real-Time Optimization (RTO) problems in order to avoid decisions that are optimal w.r.t. the model but not for the process. Among them, Modifier Adaptation (MA) methods have gained recognition as they are able to deal with both structural and parametric uncertainties while not requiring any model update, but only incorporating iteratively corrections in the on-line optimization problem obtained from measurements of the process steady-state. Nevertheless, this implies a long convergence time and the number of constraints affected by uncertainties can also limit the practical applicability of the approach. This talk presents some approaches oriented to avoid these inconveniences and, in particular, one method to speed up the convergence to the optimum by using transient information of the process. This technique is based on a recursive identification algorithm to estimate process gradients from transient measurements, achieving the plant optimum faster than traditional MA techniques. The methods are illustrated with several realistic examples covering the operation of a depropanizer distillation column and a natural gas pipeline.

 

Miroslav Fikar (Slovak University of Technology in Bratislava, Slovakia)

Miroslav Fikar

Title: Optimal control of membrane processes

Abstract:

Membrane processes stand for an emerging technology in chemical and bioprocess industry used both in production and down-stream processing. They receive growing attention mainly due to reduced energy demands and higher efficiency of the achieved separation or processing goals. These systems, however, did not receive much attention from process optimization community and that is why they provide many opportunities. The lecture will discuss membrane processes and models in details, their common features, and principles. The existing industrial operation and control will be explained. Optimal control theory will be employed to improve the existing state of control and to propose new control strategies. Simulation and experimental case studies will demonstrate the advantages of the proposed approach.

CV:

Prof. Miroslav Fikar received his ME and Ph.D. degrees in chemical engineering from the Slovak University of Technology in Bratislava in 1989 and 1994, respectively. He has stayed with the Faculty of Chemical and Food Technology STUBA where he is currently institute director. He was Postdoc Fellow in Nancy, France, Alexander von Humboldt Fellow in Bochum, Germany and has spent several stays in Denmark, Germany, France, and Switzerland. His current research interests include optimal control, MPC, and chemical process control. He is co-author of 2 international monographs, 80 journal papers, more than 200 peer-reviewed conference publications. He has been involved in several international projects, recently he acted as training manager in FP7 MSCA ITN project TEMPO.

 

Rolf Findeisen (Otto-von-Guericke University Magdeburg, Germany)

Rolf Findeisen

Title: Predictive, contract based control for connected autonomous systems

Abstract:

Due to rapid advancements in communication and information technology systems increasingly exchange information with neighboring systems or systems in the cloud. Examples are smart grids, production systems or multi-modal transportation systems. While the interconnections provide many fascinating possibilities, they also pose several challenges. What are suitable scalable control and monitoring technologies, which facilitate autonomy, while leading to an overall optimal behavior. We present contract based predictive control strategies which allow a modular design of complex systems. These can be complemented by learning strategies to cope with changing and uncertain system dynamics.

 

Ali Mesbah (University of California, Berkeley, USA)

Ali Mesbah

Title: Generalized chaos expansions with arbitrary probability measures for uncertainty quantification of stochastic nonlinear systems

Abstract:

We use the concept of generalized chaos expansion (GCE) to present a computationally efficient method for uncertainty quantification of nonlinear systems with arbitrary probabilistic uncertainties. GCE involves two key steps: (1) defining an orthonormal basis with respect to the arbitrary probability measures of uncertainties, and (2) determining the expansion coefficients. Here we present an efficient algorithm for numerically solving these two problems. The proposed algorithm only requires moments of the uncertainties to be known, alleviating the need for knowing the full distribution of uncertainties in closed-form. Leveraging moment-matching optimization, a “quasi-optimal” cubature rule will be presented for accurate estimation of the expansion coefficients based on a small number of deterministic simulations, which substantially reduces the computational cost of the method. We demonstrate the application of the proposed GCE method for the design and performance verification of MPC and optimal experiment design problems for nonlinear systems with arbitrary uncertainties.

CV:

Ali Mesbah is Assistant Professor of Chemical and Biomolecular Engineering at the University of California at Berkeley. Before joining UC Berkeley, he was a senior postdoctoral associate at MIT. He holds a Ph.D. degree in systems and control from Delft University of Technology. He is a senior member of the IEEE Control Systems Society and AIChE. He is on the IEEE Control Systems Society conference editorial board as well as the editorial board of IEEE Transactions on Radiation and Plasma Medical Sciences. He is the recipient of the AIChE's 35 Under 35 Award in 2017, the IEEE Control Systems Outstanding Paper Award in 2017, and the AIChE CAST W. David Smith, Jr. Graduation Publication Award in 2015. His research interests are in the areas of optimization-based systems analysis, fault diagnosis, and predictive control of uncertain systems.

 

Jie Lu (ShaghaiTech University, China)

Jie Lu

Title: Fenchel dual gradient methods for distributed convex optimization over time-varying networks

Abstract:

In the large collection of existing distributed algorithms for convex multi-agent optimization, only a handful of them provide convergence rate guarantees on agent networks with time-varying topologies, which, however, restrict the problem to be unconstrained. Motivated by this, we develop a family of distributed Fenchel dual gradient methods for solving constrained, strongly convex but not necessarily smooth multi-agent optimization problems over time-varying undirected networks. The proposed algorithms are constructed based on the application of weighted gradient methods to the Fenchel dual of the multi-agent optimization problem, and can be implemented in a fully decentralized fashion. We show that the proposed algorithms drive all the agents to both primal and dual optimality asymptotically under a minimal connectivity condition and at sublinear rates under a standard connectivity condition. We also derive bounds on the convergence rate and the suboptimality when the dual gradient is inexactly evaluated at each iteration.

CV:

Jie Lu received her B.S. in Information Engineering from Shanghai Jiao Tong University, China in 2007 and her Ph.D. in Electrical and Computer Engineering from the University of Oklahoma, USA in 2011. From 2012-2015 she was a postdoctoral researcher with the Automatic Control Lab, ACCESS Linnaeus Centre at KTH Royal Institute of Technology, Sweden and with the Department of Signals and Systems at Chalmers University of Technology, Sweden. Since 2015 she has been a tenure-track assistant professor in the School of Information Science and Technology at ShanghaiTech University. Her research interests include distributed optimization algorithms, large-scale optimization, multi-agent coordination, and networked dynamical systems.

 

Gabriele Pannocchia (University of Pisa, Italy)

Gabriele Pannocchia

Title: Optimization based control in the presence of uncertainties: offset-free design principles

Abstract:

Optimization based control strategies represent a general framework of numerical methods in which a (often deterministic) model of the system under consideration is deployed to achieve high-level goals (e.g., minimization of energy consumption, emission of pollutants, maximization of throughput, etc.) as well as more specific control tasks (e.g. product quality control), while respecting a number of constraints arising from physical, safety or performance limits. Feedback is necessary to reduce the effect of disturbances and to cope with unavoidable modeling errors. Nonetheless, the way in which feedback is used to achieve offset-free tracking in the presence of persistent errors or disturbances appears to be often a question of personal preference among possible different methods. The general goal of this talk is to describe, in a tutorial way, this aspect of controller design, which is often overlooked in academic papers but is often fundamental for successful implementation. After a preliminary introduction to linear offset-free Model Predictive Control (MPC) design principles based on disturbance models, explaining in detail how the integral action is achieved in spite of modeling errors, the talk will comprise three parts. 1. We propose a novel, observer-based Internal Model Control (IMC) structure which extends the simple IMC design principles to integrating and unstable plants, showing the conditions for internal stability and offset-free property. A connection with the Youla-Kucera parameterization is also established as a special case. 2. We show that several known alternative offset-free MPC algorithms (using velocity models) are special cases of the general disturbance models. 3. We extend the concepts of offset-free estimation to nonlinear tracking MPC and for the design an economic MPC algorithm that is able to cope with persistent errors while still achieving the optimal ultimate economic performance. In each part, extensive application results are presented to show the benefits of offset-free algorithms over standard ones, and to clarify misconceptions and design errors that can prevent constraint satisfaction, closed-loop stability, and offset-free performance.

CV:

Gabriele Pannocchia received the Ph.D. degree in Chemical Engineering from the University of Pisa (Italy) in 2002, where he currently is Associate Professor. He held a Visiting Associate position at the University of Wisconsin - Madison (WI, USA) in 2000/2001 and in 2008. Dr. Pannocchia is author of more than 100 papers in international journals, book chapters and in proceedings of international conferences. Dr. Pannocchia is Senior Editor for the Journal of Process Control, Associate Editor of Automatica, and in the Editorial board of Processes. He is Vice-Chair for Education of the IFAC TC 2.4 (Optimal Control). Dr. Pannocchia was IPC co-chair of the IFAC Symposium DYCOPS 2013 held in Mumbai (India), Area Co-Chair/Associate Editor in IFAC DYCOPS 2016, IFAC World Congress 2017, IFAC NMPC 2018, Control 2018. He has been keynote speaker in several international congresses (IFAC DYCOPS 2010, IFAC NMPC 2015, IFAC DYCOPS 2016). His research interests include: model predictive control systems, process simulation and optimization, numerical optimization, multivariable systems identification and performance monitoring, optimal planning and control for robotic systems.

 

Radoslav Paulen (Slovak University of Technology in Bratislava, Slovakia)

Radoslav Paulen

Title: Bi-level optimization approaches to model-based design of experiments

Abstract:

Methodologies for model-based optimal design of experiments (DoE) include mature techniques to determine the best experimental conditions for conducting an experiment so that the a posteriori data analysis yields the best (confidence) interval estimate of the unknown parameters of the fitted model. The majority of developments and applications of DoE considers maximum-likelihood estimation using least-squares approach. The optimal design is achieved by proper excitation of the system using available (experimental) degrees of freedom. Unlike the commonly used approaches, which base the DoE procedure upon linearized confidence regions (CRs), we explore a path, where we explicitly consider exact CRs in the DoE framework. As in the classical approach to design of experiments, an essential part of the solution procedure is to approximate the resulting joint-confidence region. We use a methodology for a finite parameterization of exact CRs within the DoE problem. The employed techniques formulate the DoE problem as a finite-dimensional mathematical program of bi-level nature. We study various DoE criteria and compare the resulting optimal designs with the commonly used linearization-based approaches. We also provide an extension towards the design of experiments in the framework of bounded-error (guaranteed) parameter estimation.

CV:

Radoslav Paulen received a master degree in Chemical Engineering and Process Control from Slovak University of Technology in Bratislava, Slovakia in 2008 and the doctoral degree and in Process Control from Slovak University of Technology in Bratislava, Slovakia in 2012. He was a visiting researcher at Ecole Nationale Superieure des Industries Chimiques of University of Lorraine in Nancy, France in 2011 and a visiting researcher at Imperial College London, United Kingdom in 2012. He was a postdoctoral researcher in the group of Process Dynamics and Operations in the Department of Chemical Engineering at TU Dortmund from 2012 to 2017. Presently he holds a position of Associate Professor at Slovak University of Technology in Bratislava. His publication records include 13 journal papers and 50 peer-reviewed conference publications. He has been a researcher and a member of consortium in several EU-funded projects, ERC Advanced Grant MOBOCON, FP7-ICT projects DYMASOS, HYCON-II, and CPSoS.

Sigurd Skogestad (Norwegian University of Science and Technology, Norway)

Sigurd Skogestad

Title: An overview and evaluation of approaches for online process optimization

Abstract:

This paper discusses approaches to online process optimization, where the objective is usually to minimize an economic cost. Steady-state real-time optimization (RTO) has been around for more than 25 years, but still it's not much used in practice. One reason is the steady-state wait time, because one has to wait for a new steady state before the process is reoptimized. In the talk a number of alternative approaches are discussed, given here in the order from most complex (and model-based) to most simple (and data based): 1. Economic (nonlinear) model predictive control (EMPC) 2. Dynamic real-time optimization (DRTO) 3. Hybrid RTO = Steady-state RTO with dynamic model update (new method) 4. Feedback-based Hybrid RTO (new method) 5. Conventional steady-state RTO 6. Self-optimizing control 7. Extremum seeking control, and similar feedback approaches based on measuring the cost such as NCO tracking and hill-climbing. Except possibly for EMPC, the optimizer sends setpoints to a control layer, which could be a PID layer or MPC. Dynamic stability issues are then handled by the lower control layer. In the conventional steady-state RTO, one must wait for the process to reach steady-state before updating the model using "data reconciliation". In the hybrid RTO approach, we solve the same steady-state optimization problem as in traditional Steady-state RTO, but instead of a steady-state model update, we use dynamic model adaptation with use of transient measurements, for example, using an extended Kalman Filter. This avoids the steady-state wait time. In the feedback-based hybrid RTO approach, we don't solve the steady-state optimization problem numerically as in conventional RTO, but instead the steady-state gradient is estimated by linearizing the nonlinear dynamic model around the current operating point. The gradient is controlled to zero using standard feedback controllers, for example, a PI-controller. In case of disturbances, the proposed method is able to adjust quickly to the new optimal operation. The advantage of the feedback RTO strategy compared to standard steady-state real-time optimization is that it reaches the optimum much faster and without the need to wait for steady-state to update the model. The advantage compared to dynamic RTO and the closely related economic NMPC, is that the computational cost is much reduced and the tuning is simpler. It is significantly faster than classical extremum-seeking control and does not require the measurement of the cost function and additional process excitation.

CV:

Sigurd Skogestad is a professor in chemical engineering at NTNU. He worked from 1980 to 1983 in industry (Norsk Hydro) in the areas of process design and simulation. Moving to the US, he received the Ph.D. degree from the California Institute of Technology in 1987. He is a worldwide expert in the area of process control with about 150 journal publications and a current H factor of 28 (2009). He is the principal author together with Ian Postlethwaite of the very poular book "Multivariable feedback control" published by Wiley in 1996 (first edition) and 2005 (second edition). The goal of his research is to develop simple yet rigorous methods to solve problems of engineering significance, including biological problems. Research interests include the use of feedback as a tool to (1) reduce uncertainty (including robust control), (2) change the system dynamics (including stabilization), and (3) generally make the system more well-behaved (including self-optimizing control).

 

Mario E. Villanueva, Boris Houska (ShaghaiTech University, China)

Mario Villanueva, Boris HouskaBoris Houska

Title: Set-based methods for robust control

Abstract:

To be updated.