Market variation such as changing utility price or demand can lead to non-static operation of chemical plants, such as in cases where new objectives must be met, or varying operating conditions can be taken advantage of to increase operational profit. Integrated design and control (where both the design and operation are considered simultaneously when the design of the plant is being formulated) can be used to address the challenges that plants that operate under uncertainty may face. When considering demand uncertainty for a plant, not only must the different realizations of demand be considered, but how the plant transitions from one demand to another is also of interest.
Ideally when transitioning, the plant must quickly and feasibly transition from one operating point to another. In the first study of this thesis we consider the benefit of taking into account the demand transitions a plant must undergo; compared to only considering the final operating states. In this study we examine an air separation unit (ASU) since in industrial practice ASUs can often be subject to demand uncertainty. Assessing the impact that an ASU design has on its dynamic response characteristics motivates us to include plant dynamics in the design formulation. Using a two-stage stochastic optimization framework, the optimal design parameters are found for a nitrogen plant operating under uncertain demand. Three design paradigms are explored and compared - a nominal steady-state design, a flexible design that maintains steady-state feasibility, and a dynamically operable design that enforces feasibility of dynamic transitions. The designs obtained are subjected to random demand changes to evaluate the expected economic return under transitions not directly designed for. From this study we see that when the dynamically operable design and the flexible design are both subject to dynamic transitions, the dynamically operable design in certain cases can provide more economic benefit during transition and at the final steady state.
In the second study of this thesis we address an important issue that arises when uncertainty in the plant operation is captured by utilizing two-stage stochastic optimization. If the optimization formulation can see the uncertainty profile in its entirety, then it can make control moves and design decisions based on the fact that the optimization problem knows what operating conditions it transitions between. This may not however be a realistic assumption to operate under, as future uncertainty is unknown. Given lack of foresight into future uncertainties it is logical to currently operate at the the optimal steady-state. This poses a bilevel problem for the design and control of a plant because the feasible region is determined by an inner optimization. In this study the Karush-Kuhn-Tucker conditions are incorporated into a two-stage stochastic optimization formulation for a dynamic model of chemical plant to generate a design. This design is then compared to a design generated from an optimization formulation where future knowledge of uncertainty is assumed and it is seen that the former design can provide greater economic benefit. / Thesis / Master of Applied Science (MASc)
Identifer | oai:union.ndltd.org:mcmaster.ca/oai:macsphere.mcmaster.ca:11375/24011 |
Date | January 2019 |
Creators | Jaydeep, Rohil |
Contributors | Swartz, Christopher, Chemical Engineering |
Source Sets | McMaster University |
Language | English |
Detected Language | English |
Type | Thesis |
Page generated in 0.0021 seconds