PC21 – Plenary Lectures

Plenary Lectures:

Semi-Plenary Lectures:

Plenary Lecture I

Implementing Optimization in the Process Industry

by Cesar de Prada (University of Valladolid, Spain)

ABSTRACT: Optimizing the way products are produced is a clear aim in industry that coincides with the targets of the Industry 4.0 initiative and the current digitalization trends. The topics are not new to the process industry and, nowadays, advances in many fields, from computing to methods and algorithms, open the door for successful implementations. This talk deals with the problems associated to the implementation of optimization-based systems in the process industry. It reviews several approaches for optimal process operation related to Real-Time Optimization (RTO) and Model Predictive Control (MPC). One of the main concerns when applying these techniques to processes is the presence of a process-model mismatch. Optimization-based decisions provide the optimum of the model but, quite often, its value does not correspond to the real process optimum, in particular in the presence of structural model errors. Several approaches, such as Modifier Adaptation (MA), have been proposed to overcome this limitation, but they present also practical problems when considering its implementation in an industrial environment. Recent developments have extended MA ideas to dynamic formulations within the MPC framework with the purpose of avoiding some of these problems. The talk presents some of them and illustrates their use in some practical cases.

Bio: Cesar de Prada is Emeritus Professor with the department of Systems Engineering and Automatic Control, in the School of Industrial Engineering, University of Valladolid, Spain. He received the Spanish Committee of Automatics award in 2016 and the ISA-Spain award to the professional 2008. He was also Manager of the Industrial and Design and Industrial Production research programmes of the Spanish Ministry of Science and Technology between 2000-04. His fields of interest focus on the control and dynamic optimization of process systems, as well as in modelling and simulation. His research topics are now focused on optimal management of large-scale systems, considering aspects such as scheduling, RTO, presence of uncertainty and hybrid continuous-discrete elements and hybrid model developments. The research combines the development of methods and algorithms with industrial applications and active participation in EU projects.

Plenary Lecture II

Digitalization Drivers and Vision in the Refining and Petrochemicals Industry

by Andras Buti (Slovnaft, a.s., Slovakia)

ABSTRACT: The main, recent Drivers and the future Vision of automation-digitalization will be presented; from an end-user perspective. The ultimate goals in our industry are:

  • to achieve best utilization of people and assets
  • to increase efficiency, productivity and safety

Independent benchmarking studies (Solomon, SPI) assess Refineries, and compare them to each other – giving an objective feedback to management and highlight areas to be improved. These studies are also important drivers of changes in our industry. The automation-digitalization goals can be splitted into two main areas:

  • industrial process control (according to the ISA95 and ISA99 layering concept). In this case, the technology (the physical – chemical process) is given. The main challenge is to improve and optimize the installed base.
  • office “workflow automation” (often also referenced to as “RPA” = robotic process automation). In this case, the process (office/business workflows) is flexible, can be changed.

Therefore, the challenge is a combination of optimizing the workflows (not a digitalization-automation task) and selecting-deploying the right digital tools for the purpose. The presentation will focus on the first area. In the Vision part, I will present our main initiatives to achieve the goals.

  • Managing legacy systems and converting silos to integrated, interconnected systems
  • IIoT, smart sensors, edge and cloud for condition-based maintenance and environment protection
  • Data, information transparency and availability for own employees, as well as for external contractors and licensors
  • “Self-service” data science and analytics
  • Unit and multi-unit optimization

Bio: Andras Buti holds an MSc in Electrical Engineering (Budapest University of Technology and Economics). He has 14 years of experience in Process Control, in Advanced Process Control and in Simulation. He leads the Process Information and Automation team in Slovnaft.

Semi-Plenary Lecture I

Lyapunov Based Adaptive Control for Varying Length Pendulum with Unknown Viscous Friction

by Milan Anderle (Czech Academy of Sciences, Czech Republic)

Best Paper Award

ABSTRACT: The paper deals with pendulum swing suppression based on controlling the pendulum length. The continues up and down movement generates the Coriolis force to damp the pendulum swing movement. The crucial difficulties of the pendulum system to be stabilized depicts in its approximate linearization at the desired equilibrium which is neither controllable, nor stabilizable. Therefore any simple controller cannot be directly applied on the system to damp the pendulum swing. The well known Lyapunov method together with the backstepping technique were successfully used to damp the pendulum swing. Moreover, by virtue of aplication of the backstepping control, the adaptive estimation of the viscous friction coefficient can be simply applied and used to improve the damping performance of the pendulum swing. The improvement of pendulum swing damping by virtue of the adaptive estimation of the viscous friction coefficient is demonstrated using simulations and real-time experiments.

Bio: Milan Anderle is research associate with the department of Control theory in the Institute of Information Theory and Automation of the Czech Academy of Sciences and also assistant professor at the Department of Instrumentation and Control Engineering, Faculty of Mechanical Engineering, Czech Technical University in Prague. He received the Ph.D. degree in the control engineering and robotics at the Czech Technical University in Prague. The topic of his PhD thesis were nonlinear control laws for underactuated biped walking robots. His research interests include nonlinear control of walking robots and applications of input shapers for control of multibody systems considering aspects of mathematical modelling, numerical simulations, real-time control and microprocessor systems, electronics.

Semi-Plenary Lecture II

Clustering-based Optimal Dynamic Pricing for Residential Electricity Consumers

by Salma Taik (Budapest University of Technology and Economics, Hungary)

Best Paper by Female Researcher Award

ABSTRACT: Electric power shortage in a residential area may occur with an increased probability if appropriate coordination mechanisms are missing. Time-of-Use (ToU) dynamical pricing has been proposed to influence the demand-side consumption to ensure a stable and optimal power system operation. This paper presents a method to find an optimal ToU electricity tariff if a single utility company (UC) provides electricity. The tariff is obtained based on the analysis of historical consumption data. First, the real consumption data is analyzed and clustered to select the possible consumer population to be targeted by ToU tariffs. Second, a simple consumer behavior model is established to predict the consumption profile changes if the ToU tariff is applied. Third, the GA-based optimization resulting in the tariff is carried out. The goal is to ensure a win-win situation for the consumers on the demand-side and the UC when the optimal ToU is employed. The effect of the dynamic pricing is demonstrated by simulating the case of one consumers category.

Bio: Salma Taik received an MSc in Electrical engineering from Tetouan Faculty of Science, Morocco, in 2016. She is currently pursuing her Ph.D. studies at the Department of Control Engineering and Information Technology at Budapest University of Technology and Economics, Hungary. Her research interests include control and optimization in a smart grid environment.

Semi-Plenary Lecture III

Reinforced approximate robust nonlinear model predictive control

by Benjamin Karg (TU Dortmund University, Germany)

Best Paper by Young Author Award

ABSTRACT: Model predictive control (MPC) has established itself as the standard advanced process control method. However, solving the resulting optimization problems in real-time can be challenging, especially when uncertainty is explicitly considered in a robust nonlinear predictive control approach. An increasingly popular alternative to avoid the online solution of the resulting optimization problems is to approximate their solution using neural networks. The networks are trained using many solutions of the MPC problem for different system states and therefore this approach is often called imitation learning. Controllers obtained via imitation learning have important drawbacks, since it is difficult to learn behaviors that are not well represented in the data and they must be redesigned from scratch when the control task changes. In this work, we show that these two drawbacks can be mitigated by combining imitation learning and concepts from reinforcement learning. The central idea is to use imitation learning as a very good initialization of a control policy that is iteratively updated using reinforcement learning, taking advantage of the fact that an explicit and differentiable expression of the approximate MPC controller is available. The efficacy of the combination of the two learning paradigms is highlighted via simulations of a semi-batch industrial polymerization reactor.

Bio: Benjamin Karg started his PhD in July 2017. His research interests are control engineering, artificial intelligence and edge computing. By exploiting the expressive capabilities of deep neural networks, complex control and decision-making algorithms can be approximated, which in return allows the deployment of said algorithms on computationally limited hardware. Currently, he is using optimization-based and probabilistic methods and ideas from reinforcement learning for the verification of learned controllers to obtain guarantees on performance and safety with respect to operational requirements and for modifying the controllers such that they satisfy the performance criteria.