Use this URL to cite or link to this record in EThOS: https://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.799416
Title: Operational optimization of crude oil distillation systems with limited information
Author: Yang, Xiao
Awarding Body: University of Manchester
Current Institution: University of Manchester
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
Abstract:
Crude oil distillation is the locomotive of refining and petrochemical industries. Due to massive throughput and energy demand of industrial crude oil distillation systems, even a minor improvement in their operations can bring significant economic and social benefits. However, there are two major practical challenges for operational optimization of crude oil distillation systems. One is that limited information of crude feed compositions is known for optimization. The other one is that it's difficult to balance accuracy, complexity and robustness of optimization models. In this work, two types of methods are proposed with different philosophies of utilizing limited information during the procedure of operational optimization. The first type of method, real-time optimization, tries to use more amount of information during optimization by parameter estimation. The second type of method, robust operational optimization, tries to use less amount of information during optimization and treats limited information as uncertainty. For real-time optimization methods, a framework to simplify rigorous models with crude feed estimation is proposed. The simplified linear models are shown to have the advantage of small size and convexity with accepted accuracy loss compared to rigorous models. Second, a model correction mechanism is proposed to further improve model accuracy and reduce mismatches between models and the process. Third, a framework to mine historical data for building data-driven models based on crude feed estimation is proposed. For robust operational optimization, a method to describe the crude feed uncertainty based on simplified linear models is proposed. Second, a framework to update both certain and uncertain parameters from schedule of crude oil operations and real-time plant measurements for online use is proposed. Third, a method to determine the best shape and size of the uncertainty set and reduce loss of optimization potential is proposed. Case studies show that both real-time optimization and robust operational optimization can help to make operational optimization decisions with limited information. Real-time optimization can be expected to obtain more optimization potentials but also takes risks of worse operating conditions or infeasible operations caused by bad parameter estimation. Robust operational optimization makes conservative optimization decisions but can provide safeguard against the assumption of perfect parameter estimation implied by real-time optimization.
Supervisor: Smith, Robin ; Zhang, Nan Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID: uk.bl.ethos.799416  DOI: Not available
Keywords: robust optimization ; real-time optimization ; operational optimization ; crude oil distillation
Share: