Economic pressures require higher flexibility in operation of oil refineries, resulting in e.g. processing varying feedstocks. Different feedstocks mean varying operating conditions in the production units. To operate efficiently, the units must be controlled and operated in an integrated fashion. High-level planning tools are usually deployed to set productions targets considering product prices and customer orders. These targets can be, however, fulfilled using various production paths due to the integrated nature of the site. The goal of thesis will be developing methodology for selecting the most energy-efficient production path in real-time. This will be done using dedicated Key Performance Indicators (KPIs). It must also be ensured that the production units fulfill these tasks. Thus, the second aim of thesis is to transform the KPIs into a set of technical variables (e.g. streams purity) in a systematic way and to introduce these values into the advanced process controllers of the production units. The considered work is carried out in a cooperation with an industrial partner. Acceptance of the proposed solution with plant managers and operators should be achieved.
As the computers and algorithms get generally faster, many new control concepts become tractable and can be developed. Set-based control is one of these, where the primary use of sets is in enveloping a space of possible evolutions of variables of a system over time. If these envelopes can be obtained in reasonable time, many properties of dynamic systems such as stability or robustness can be reasoned about. The first goal of the thesis is to build a novel type of multi-base set arithmetics that combines elements such as interval analysis, convex-set theory, and polynomial-functions theory to achieve the best trade-off between accuracy of representation and the burden associated with the underlying calculations to obtain the envelopes. The second goal of the thesis is to develop methods of synthesis of controllers that can be used for safe and reliable control of nonlinear systems. The project of the thesis will be finished with a successful demonstration of the developed techniques on a laboratory plant.
Modern control architectures comprise elements designed to identify an optimal operating regime of the system, to reduce the uncertainty in key system variables by estimation using measurements (usually corrupted by noise), and to steer the system to the optimal operation regime by dynamic adjustment of available degrees of freedom. Despite these technologies being well understood, there are challenges to be addressed e.g., increased level of system complexity. One way of addressing the challenges is to combine first-principles and data-based modeling, i.e. to use hybrid modeling. As the data-based parts of models bear a significant level of uncertainty, which propagates through the whole control system and, thus, must be considered in the control design. Using the set-based methods, control values can be found to steer the system into an optimal regime in robust fashion while avoiding any violation of restrictions imposed by production quality or safety. The goal of the thesis is thus to innovate control system architecture such that the elements of the control system can exchange information about levels of uncertainty in the signals at their output and the acceptable level of uncertainty in the input signals. The project of the thesis will be finished with a successful demonstration of the developed techniques on a laboratory plant.