DIFFER Seminar: System Identification: learning for decision and control

Abstract: 

Mathematical models are crucial in many scientific fields. Modelling that starts from first principles can then be a good start. For operational use in real practice, however, the model should be in good accordance with experimental data. Nowadays, data-driven modelling becomes more and more popular, thus using a limited amount of prior knowledge.

The aim of this presentation is to introduce methods for the determination or identification of static/dynamic models starting from limited or full prior system's knowledge, given experimental data and the model objective, and to estimate the unknown parameters. The methods will be illustrated by some simple examples and a real-life case.

Biography: 

Karel. J. Keesman is Personal Professor “Systems Theory for Sustainability” at Wageningen University. He received his Ph.D. degree for his work on set-membership identification and prediction of ill-defined systems, with application to a water quality system, University of Twente, in 1989. His main research interests focus on identification, modelling and control of uncertain dynamic systems, such as bioreactors, food storage facilities, and on biobased socio-economic/environmental systems within the Water-Energy-Material nexus with either small or big data sets. He published more than 250 papers in international journals and refereed proceedings. Since 2009 he is scientific project manager/senior advisor at Wetsus for 1 day/week.