From Real-Time Optimization Techniques to an Autopilot for Steady-State Processes
Aris Papasavvas
DOI: https://doi.org/10.48550/arXiv.2108.08715
2021-08-19
Abstract:Any industrial system goes along with objectives to be met (e.g. economic performance), disturbances to handle (e.g. market fluctuations, catalyst decay, unexpected variations in uncontrolled flow rates and compositions,...), and uncertainties about its behavior. In response to these, decisions must be taken and instructions be sent to the operators to drive and maintain the plant at satisfactory, yet potentially changing operating conditions.
Over the past thirty years many methods have been created and developed to answer these questions. In particular, the field of Real-Time Optimization (RTO) has emerged that, among others, encompasses methods that allow the systematic improvement of the performances of the industrial system, using plant measurements and a potentially inaccurate tool to predict its behaviour, generally in the form of a model. Even though the definition of RTO can differ between authors, inside and outside the process systems engineering community, there is currently no RTO method, which is deemed capable of fully automating the aforementioned decision-making process. This thesis consists of a series of contributions in this direction, which brings RTO closer to being capable of a full plant automation.
Keywords: Real-time optimization, Decision-making, Optimization, Reduced-order-model optimization, Autopilot for steady-state processes, Operational research.
Optimization and Control,Systems and Control