Achieving effective control of deaeration chemical dosing using artificial neural networks

McCairns, Michael (2018) Achieving effective control of deaeration chemical dosing using artificial neural networks. [USQ Project]


The injection of seawater into hydrocarbon bearing reservoirs is an important secondary process in the offshore oil and gas industry, however the naturally occurring dissolved oxygen in the seawater is problematic and its removal prior to injection is achieved by vacuum de-aeration supported by chemical scavenging. This research aimed to achieve effective control over the dosing of the chemical scavenger to ensure the dissolved oxygen specifications were met whilst minimising the quantity of chemicals used. Neural control systems have been shown to be more effective at controlling highly non-linear processes than standard controllers and the use of a neural controller was therefore specified as a main objective of this research.

Data from several process parameters were captured using a variety of methods, including the establishment of a wireless HART communications network. A neural NARX network was successfully trained to model the de-aeration process using this captured process data. In a novel approach, a neural Model Predictive Controller was then designed, trained and evaluated in the Simulink environment using the neural NARX as the plant model. An industry standard PID controller was used as a benchmark against which to compare the Model Predictive Controller performance.

Extensive testing of both controllers revealed that due to inconsistencies within the neural models, likely caused by small training datasets and the accumulation of prediction errors, the PID controller was able to outperform the Model Predictive Controller. The Model Predictive Controller would be able to provide functional, but not effective, control if subjected only to long duration, step type references in circumstance where steady-state error was not a significant concern.

Statistics for USQ ePrint 40643
Statistics for this ePrint Item
Item Type: USQ Project
Item Status: Live Archive
Faculty/School / Institute/Centre: Historic - Faculty of Health, Engineering and Sciences - School of Mechanical and Electrical Engineering (1 Jul 2013 - 31 Dec 2021)
Supervisors: Wen, Peng; Chin, Hing Yuen
Qualification: Bachelor of Engineering (Honours) (Instrumentation Control and Automation)
Date Deposited: 09 Sep 2021 04:36
Last Modified: 29 Jun 2023 01:54

Actions (login required)

View Item Archive Repository Staff Only