Background and activities

PhD title: Stochastic Nonlinear Model Predictive Control

PhD description:

Nonlinear model predictive control (NMPC) is a popular control method for highly nonlinear and unsteady state processes. Its main advantages are its ability to deal with constraints and strongly coupled, multi variable plants. The introduction of uncertainty in the system can however lead to sub-optimal behaviour and failures of the classical NMPC algorithm. Common uncertainties in NMPC are parametric uncertainties, model structure uncertainties, disturbances and state-estimation error, which can be found in nearly any real implementation. In robust control min-max or worst-case methods are applied to guarantee stability and recursive feasibility for which the uncertainties are given by bounded sets. Robust methods can however be overly conservative, since the probability of occurrence of different uncertainty realizations is not considered. Stochastic NMPC (SNMPC) on the other hand exploits probabilistic uncertainty descriptions, which can be used to trade-off between probabilistic constraint satisfaction and performance. SNMPC describes a general formulation and covers many different potential approaches to deal with uncertainties for MPC, including techniques from nonlinear filtering, stochastic programming and non-parametric modelling. 

Main Supervisor: Lars Imsland

Co-Supervisors:  Bjarne Foss, Marcus Reble (BASF)

Short CV: 

2015–2016 MPhil Chemical Engineering, University of Cambridge.
2011–2015 MEng Chemical Engineering, Imperial College London,
First Class Honours.

Acknowledgements:

The research is carried out as part of the European Union's Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 675215.

I gratefully appreciate BASF for hosting my placement over the
course of the Marie-Curie project