Coordinated Feedback-optimizing Control for Large Scale Processes - with applications of field-wide oil & gas production system

Abstract

This thesis proposes optimization strategies for large-scale process systems, such as those found in oil and gas production, characterized by diverse subprocesses and constraints. This can be achieved by utilizing real-time optimization (RTO), which provides optimal setpoints for the control layer. Despite its benefits, RTO is underutilized due to high model development costs, infrequent updates, and the need for specialized skills. As an alternative, this thesis advocates for a pragmatic approach using common tools like PID controllers and selectors, complemented by small-scale numerical solvers (if necessary), to enhance practical implementation feasibility. First, the thesis addresses active constraint switching by suggesting a primal-dual framework that allows automatic constraint switching with a fixed control structure. For processes that require tight constraints, the idea of integrating override constraint controllers is proposed with auxiliary constraints to preserve the automatic switching. The second part starts by presenting a dual-based distributed feedback-optimizing system that decomposes large-scale systems into manageable subsystems. This facilitates system-wide optimal performance through a coordinating subsystem. This approach relies on accurate real-time gradient estimations. For cases with non-performing local gradient estimations, this thesis proposes systematic pairing procedures for an override controller to minimize economic losses. For processes with more available inputs and a limited input rate, a multi-input override control procedure is proposed. For processes with total input constraints, a primal-based distributed feedback-optimizing system with compensator is suggested. Experimental validation on a lab-scale gas-lift rig demonstrates these approaches’ effectiveness. The final part addresses practical issues by proposing a decentralized graph-based primal-based distributed feedback-optimizing system for marginal offshore oil fields to enhance adaptability and eliminate the coordinator. A gas-lift system model is also extended, and this thesis suggests self-optimizing control for cases with limited sensor issues. Overall, this thesis emphasizes the need for unique optimizing-control strategies to address specific challenges for distinct problems.”