By Ayo Onikoyi

In a landmark contribution to petroleum engineering, Nigerian researcher Prosper Nekekpemi has developed a machine learning model that significantly improves the prediction of bubble-point pressure in crude oils.

The study, presented at the prestigious 2024 Offshore Technology Conference (OTC) in Houston, offers a more accurate and cost-effective alternative to traditional prediction methods.

Bubble-point pressure is the point at which gas first separates from oil in a reservoir—a critical parameter in oil production planning. However, conventional methods for determining this property often involve costly and time-consuming laboratory experiments. Nekekpemi’s model harnesses advanced machine learning algorithms to offer a faster, more scalable solution.

The research evaluates several algorithms—including Decision Tree, Random Forest, Support Vector Regression (SVR), and Gradient Boosting—using a dataset of 776 global crude oil samples. The Gradient Boosting model emerged as the top performer, achieving a high coefficient of determination (R² = 0.924) and a low root mean square error (RMSE = 364.03).

Findings by this news medium indicated that this model surpasses the accuracy of several widely used empirical correlations in the field.

“The novelty of our approach lies in its ability to learn from diverse and complex data,” Nekekpemi said in an interview.

“Unlike conventional models that struggle with non-linear relationships, our machine learning models adapt and improve as more data becomes available.”

Drawing on data from North America, the Middle East, and Nigeria’s Niger Delta, the study reflects a global perspective. “Incorporating data from the Niger Delta was intentional,” Nekekpemi explained. “It ensures the model is inclusive and applicable to regions that have historically lacked accurate predictive tools.”

Nekekpemi, a trained petroleum engineer with dual Master’s degrees in Engineering and Informatics (Data Science concentration), has a proven track record in artificial intelligence and reservoir modelling. His academic and professional journey spans Chevron Nigeria, TC Energy in the U.S., and deepwater exploration roles in Louisiana.

He has also received numerous awards, including the SPE Imomoh Scholarship and the 2025 Excellence in Energy & Sustainability Award from the National Institute of Professional Engineers and Scientists (NIPES), where he is a Fellow. His leadership roles in the Society of Petroleum Engineers (SPE) and mentoring activities in Africa and North America highlight his commitment to shaping future engineers.

The breakthrough is especially relevant as energy producers seek ways to enhance production efficiency and reduce carbon emissions. By improving reservoir simulation and reducing dependence on laboratory experiments, the model supports both cost savings and environmental responsibility.

“The oil and gas industry is undergoing a digital transformation,” Nekekpemi noted. “Machine learning is not just a trend—it’s a necessary evolution. We are entering an era where data can drive smarter, more sustainable decisions.”

His study goes beyond just prediction. It also evaluates the relative importance of each variable affecting bubble-point pressure. According to the Gradient Boosting model, solution gas-oil ratio (GOR) is the most influential factor, followed by gas gravity, reservoir temperature, and oil gravity.

Comparisons with older empirical models, such as those by Standing and Vazquez & Beggs, demonstrate that machine learning offers a more generalisable and accurate approach. “The models we evaluated were built decades ago, often with limited regional data,” said Nekekpemi. “Our approach benefits from broader datasets and contemporary computational power.”

Experts in the field are already taking note. His work has implications for improved reservoir management, production forecasting, and real-time decision-making—especially in data-scarce regions.

Looking ahead, Nekekpemi plans to expand the research by integrating the model into intelligent reservoir monitoring systems. He also aims to apply similar techniques to other oilfield challenges, including CO₂ sequestration and geothermal energy production.

“This is only the beginning,” he said. “Our goal is to create accessible, AI-powered tools that will redefine how we understand and manage subsurface resources. That’s how we future-proof energy.”