Loading...
Optimisation of produced water deoiling in hydrocycloneinduced gas flotation system: a case of machine learning
Author(s)
César, Sandro Duarte
Date Issued
2026
Type
master thesis
Publisher
Cape Peninsula University of Technology
Abstract
Produced water represents the largest liquid waste stream generated by offshore oil and gas operations, often exceeding hydrocarbon output in mature fields. Its complex composition, including dispersed and emulsified oils, creates significant environmental and operational challenges. To comply with stringent discharge standards, offshore platforms employ multistage treatment systems such as Hydrocyclones and Induced Gas Flotation (IGF) units. However, their performance is highly sensitive to fluctuating inlet conditions, equipment aging, and chemical dosing variability, making consistent regulatory compliance difficult. This study developed an integrated, Machine Learning–aided digital framework to model, predict, optimise and intelligently control offshore produced water deoiling systems, focusing on Hydrocyclone–Induced Gas Flotation (HIGF) units. Using operational data from offshore facilities, ensemble learning models, particularly Extreme Gradient Boosting (XGBoost) and Gradient Boosting, effectively captured nonlinear process dynamics, achieving predictive accuracies of R² = 0.78 for the Hydrocyclone and R² = 0.59 for the IGF, with corresponding RMSE values of 5.7 ppm and 3.9 ppm. These models identified key operational drivers, including inlet oil-in-water concentration, clarifier dosage and differential pressures, providing actionable insights for optimisation. By integrating the predictive models within a Monte Carlo simulation framework, the research quantified compliance risk and performance variability under uncertain conditions. The conceptualisation of a digital twin further demonstrated the feasibility of real-time monitoring, adaptive optimisation and control under dynamic offshore environments. Overall, this work provides a validated foundation for intelligent, data-driven produced water management, advancing environmental compliance, operational efficiency and sustainability in offshore oil and gas production.
Additional information
Thesis (DEng (Chemical Engineering))--Cape Peninsula University of Technology, 2026
File(s)![Thumbnail Image]()
Loading...
Name
Cesar, S_208184732 (1) (1).pdf
Size
14.91 MB
Format
Adobe PDF
Checksum
(MD5):8156a7c80117fafeba03e41a2ec22d0d