Date of Award

Spring 5-2015

Embargo Period

1-6-2016

Degree Type

Dissertation

Degree Name

Doctor of Philosophy (PhD)

Department

Chemical Engineering

Advisor(s)

Larry Biegler

Abstract

Over the past thirty years, flowsheet optimization methods have evolved from “black box” approaches to sophisticated equation-oriented methods for simultaneous flowsheet convergence and optimization. This thesis explores the next generation of flowsheet optimization tools that leverage completely open models (with exact first and second derivatives) and utilizes start-of-theart nonlinear programming (optimization) solvers. A five part framework is proposed in this thesis: 1. Embedded cubic equation of state thermodynamic models with complementarity constraints to accommodate vanishing and reappearing phases 2. Simultaneous heat integration and process optimization using the pinch location method 3. Aggregate short-cut and rigorous tray-by-tray distillation models 4. Steam cycle equipment (e.g., turbine) and boiler models 5. Trust region optimization algorithm to incorporate models with expensive derivatives into the equations-based framework A systematic initialization routine based on model refinement and multistart procedure are also presented as practical alternatives to global optimization. Complementarity constraints are used throughout the framework to model switches, such as vanishing phases. Degeneracy Hunter, an algorithm that identifies irreducible sets of degenerate constraints (i.e., constraints with a rank deficient Jacobian) is proposed and used to refine the models. The framework is demonstrated in a series of case studies related to the design of oxycombustion power systems with CO2 capture. Two case studies focus on the simultaneous optimization of gases separation systems and their accompanying multistream heat exchangers. In one of these case studies, the optimization procedure identifies common air separation unit configurations with comparable specific energy requirements to industrial designs. The framework is also used to optimize regenerate Rankine cycles, where steam flowrates from nine extraction points for boiler feedwater heating are considered as optimization variables. This allows for waste heat from compression to the completely integrated into the steam cycle. Steam table lookups (without derivatives) are incorporated using reduced models and a trust region optimization algorithm.

Share

COinS