For Emuobosa Patience Ojoboh, SPE, the future of reservoir engineering lies at the intersection of physics, geoscience, and artificial intelligence. As a PhD researcher in petroleum engineering at The University of Tulsa, she is tackling one of the industry's long-standing challenges: making time-lapse (4D) seismic data a practical tool for assisted history matching.
For decades, production data has been the foundation of reservoir history matching, providing valuable insight into how reservoirs perform over time. Yet production data offers only part of the story. Engineers still have limited visibility into what happens between wells, where much of the reservoir remains hidden. While 4D seismic surveys have the potential to reveal those changes, integrating them into routine history-matching workflows has traditionally required complex seismic inversion and petro-elastic modeling, making the process computationally intensive and introducing additional uncertainty.
Ojoboh's research aims to change that.
"Production data tells you what's happening at the wells, but the reservoir between them stays a black box," Ojoboh says. "Seismic data can fill that gap, I wanted to build something that actually uses it, instead of leaving it on the table because the physics is hard."
Working within The University of Tulsa Petroleum Reservoir Exploitation Projects (TUPREP) research consortium, she is developing deep-learning workflows that seek to learn the relationship between seismic responses and reservoir behavior directly. By reducing reliance on conventional inversion-based workflows, her research could help make seismic-assisted history matching faster, more efficient, and more practical for reservoir engineers.
A key part of her work involves building an independent SIM2SEIS workflow that converts reservoir simulation models into synthetic seismic data. Rather than relying entirely on existing software, the workflow reconstructs each stage of the forward-modeling process, from petro-elastic modeling to seismic convolution, and validates each stage directly against real field production data rather than assumed parameters. Ojoboh has already processed the workflow across more than 100 realizations of the Brugge benchmark field, building the large, physics-validated data set needed to train deep-learning models capable of reading reservoir behavior straight from seismic signals. The result is a flexible framework that gives researchers greater control over the underlying physics while generating realistic training data at scale.
The research is being conducted under the supervision of Mustafa Onur, SPE, director of TUPREP and McMan Professor and Chairman of the McDougall School of Petroleum Engineering, in collaboration with fellow PhD researcher Diego de Miranda Gomes. Together, their work supports TUPREP's broader mission of advancing reservoir characterization, assisted history matching, production optimization, and uncertainty quantification through innovative computational methods.
The broader objective is to extend these workflows to field-scale reservoirs, where seismic data can help engineers better understand fluid movement and reduce uncertainty in reservoir models.
As artificial intelligence becomes increasingly integrated into the energy industry, Ojoboh believes its greatest value lies in strengthening, not replacing fundamental reservoir engineering. By combining physics-based understanding with modern machine-learning techniques, her research is helping build seismic-assisted history-matching workflows that give engineers a clearer view of the reservoir between wells.
