- 238 views
- 285 downloads
Process Data Rectification and Process Optimization with Uncertainty
-
- Author / Creator
- Yuan, Yuan
-
Process measurements collected from daily industrial plant operations are essential for process control and optimization. However, due to various reasons, the process data are always corrupted by errors, so that process model constraints representing the mass balance and energy balance are not satisfied with measured data. If the data with errors are used in process control or optimization, the results may not be appropriate for the system, and cannot achieve the desired target, or even worse, it may be hazardous to the system and even cause damage and break-down. Random errors and gross errors are two major sources of errors, and techniques are needed to detect and eliminate the errors from the measurements to obtain clean data for further use. Even after the processing of the data, there remain some uncertainties in the data. After passing the data to optimization problems, due to the existence of uncertainty in the data, deterministic optimization formulation can no longer be utilized in order to avoid suboptimal or infeasible solutions. Optimization with uncertainty becomes an important topic in both research and applications. In this thesis, a technique is first developed to detect the gross errors and reconcile the data simultaneously to remove the errors from the data. A hierarchical Bayesian algorithm is used to formulate a unified framework to detect the gross errors, estimate the magnitude of the gross errors, determine the covariance matrix of the random errors, and reconcile the data. Among various approaches for optimization with uncertainty, chance constraint problem is a natural way to quantify the reliability of the solutions by setting a restriction on the level of the probability that the constraints are satisfied. In the case that multiple constraints should be satisfied simultaneously, joint chance constraint is appropriate to model the uncertainties. However, joint chance constraint problem is generally intractable and a variety of methods are available to approximate it into tractable forms. Robust optimization with the distribution-free property is an approach with computational advantage. In this thesis, a novel framework is proposed to approximate the joint chance constraints using robust optimization and improve the approximation results using a two-layer algorithm to optimize two types of important variables. There are always correlations between different measurements or data. It is necessary to consider the correlations in the data uncertainty. In this thesis, the robust optimization formulation based on the uncertainty set incorporating correlations of uncertainties is studied. Furthermore, nonlinearity is commonly seen in practical process models. This thesis develops a novel robust optimization framework to consider the uncertain nonlinear optimization problems. The thesis provides practical applications as well. An economic optimization problem is investigated for steam generation and water distribution for SAGD (steam-assisted-gravity-drainage) process. The uncertainty in oil production capacity is considered and the proposed robust optimization algorithms are utilized to solve the optimization problems that contain uncertainty.
-
- Graduation date
- Fall 2017
-
- Type of Item
- Thesis
-
- Degree
- Doctor of Philosophy
-
- License
- This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.