Publications

Slugging presents significant operational challenges in offshore gas production, mainly when fluids from multiple wells are transported from an offshore Wellhead Platform (WHP) to a Central Processing Plant (CPP) through a single flowline. Currently, slugging in the pipeline from WHP Charlie (WHPC) to CPP at NB Field is detected through two indicators: pressure increases (backpressure) at the sand separator on WHPC and low liquid production rate at CPP. These indicators prompt field production engineers to perform immediate treatment actions, such as opening idle wells to alleviate the slugging effects. However, this process is reactive and relies heavily on human decision-making after slugging occurs, which leads to unstable flow rates and ultimately affects reserve estimation.

This study presents a data-driven approach to enhance early detection and management of slugging by implementing machine learning algorithms. The developed models leverage real time operational data, including pressures and production rates from WHPC and CPP, to identify early signs of slugging. Several models have been constructed for this study, including a Random Forest Regression model, Extreme Gradient Boosting, Light Gradient Boosting, and many other machine learning algorithms. The result shows that a Random Forest Regressor model can achieve a coefficient of determination (R²) value of 0.98 and a mean absolute error (MAE) of 9.8, which shows the model's ability to predict time to slugging events.

Implementing this model provides substantial operational benefits. By accurately predicting slugging events, this data-driven approach enables proactive management to stabilize production flow. With 9-10 hours lead time from the Random Forest Regression model's MAE, operators can take preemptive actions—such as adjusting flow rates or activating idle wells—to minimize or even mitigate slugging events more effectively. This predictive capability also enhances reserve estimation by reducing production losses due to slugging and leading to improved reserve management.