Advanced Data-Driven, Artificial Intelligence Models Improve Oil-in-Place Estimation in Niger Delta

Accurately estimating the amount of oil contained in underground reservoirs remains one of the most persistent challenges in petroleum engineering, particularly in complex formations like Nigeria’s Niger Delta.

Conventional volumetric methods often rely on averaged reservoir data, introducing significant uncertainty at the appraisal stage.

To address this limitation, Imaobong Udoh Tom, a petroleum engineer, developed innovative computational models that combine traditional engineering techniques with modern data-driven and artificial intelligence approaches to improve the accuracy of oil in place (OIP) estimations.

Her study focused on three Niger Delta fields—Philus, Otu, and an Unnamed field—to evaluate and enhance existing estimation methods.

“To assess the reliability of volumetric estimations, we integrated Monte Carlo simulations to establish proved, probable, and possible reserves,” Tom explained. The results revealed that OIP estimates were lower than proved reserves but higher than probable and possible reserves, with uncertainties ranging from 18% to 20%.

Tom sought to reduce this uncertainty by developing alternative models based on the design of experiments (DOE) and artificial neural networks (ANN). These models provided more refined predictions of OIP across the studied fields.

Her DOE model closely mirrored volumetric estimates, with R² of 0.9971, demonstrating near-perfect alignment with field data. The ANN model achieved R² of 0.9241, incorporating machine learning to capture complex variable interactions.

The study highlighted the contribution of key reservoir parameters: initial water saturation, porosity, oil column height, and reservoir area. “Our models show that machine learning can complement traditional engineering methods, providing higher confidence in reservoir evaluations,” Tom noted.

For the Philus field, the DOE model estimated 0.2105 billion STB, while the ANN model predicted 0.2556 billion STB. Similar improvements were observed for Otu and the Unnamed field.

Tom emphasised that these methods can serve as alternative tools for oil companies operating in the Niger Delta, particularly where data is sparse or uncertain.

She conducted this research during her Bachelor of Science program at the University of Uyo, where she graduated with top honors. Her work has attracted attention at conferences in Nigeria and internationally, and has since been published in the International Journal of Scientific and Engineering Research, highlighting her early contribution to the field.

Beyond publication, Tom’s research reflects the growing integration of data analytics and artificial intelligence in petroleum engineering. These tools enhance reservoir characterisation and improve decision-making during field development.

“Adopting data-driven models in reservoir evaluation allows for better risk management and more informed investment decisions,” Tom said, highlighting the practical significance of her study.

With rising global energy demands, Tom’s contributions demonstrate the role of advanced computational methods in maximising resource recovery while reducing operational uncertainty.

As energy companies look to optimise production, researchers like Tom are proving that combining engineering fundamentals with modern data science and artificial intelligence technologies can yield substantial benefits.