Resource optimization is the key challenge facing oil and gas companies today. During exploration and other upstream activities, optimizing resources depends on accurate models that illuminate the subsurface potential.
At Galp, geoscientists create such models by analyzing 3D seismic echoes, notes from scientists, technical reports and academic papers.
Time is critical, but because the data is terabyte-class and the insights hidden in unstructured documents, the analysis is complex and time consuming. Plus, the results can be inconsistent due to the scientists’ inevitable subjectivity.
AI powers a new way of oil and gas exploration decision making
To improve the process, we worked with IBM Research in Brazil to create a GeoScience Advisor that combines the digital with the human. The advisor uses AI to collect and organize the data and analyze it using knowledge drawn from our geoscientists. Then, it offers suggestions along with supporting evidence to inform exploration decisions.
The advisor accelerates the analytical process and improves model accuracy. Its insights can help Galp optimize assets while lessening the risks of missing rich deposits or drilling dry wells.
Our innovation team had long wanted to apply AI to exploration analytics, but some colleagues were skeptical that AI could do the job. Their main concern was that analyzing diverse and huge volumes of structure and unstructured data to create subsurface models was too complex for AI. IBM Research gave us a different perspective.
IBM Cloud will provide worldwide access to the advisor
Intrigued by the concept, IBM researchers worked diligently to understand the complexities of subsurface characterization. We collaborated as one team, Galp providing domain expertise aided by IBM’s technology prowess. After a number of sessions to brainstorm solution requirements and resolve technical issues, we agreed to a three-year R&D project to develop the advisor.
The project’s initial focus is the prolific Santos basin in Brazil. The work is so promising that it is funded by Brazil’s National Agency of Petroleum, Natural Gas and Biofuel. At first, development proceeded using on-premises computing clusters. Now that the advisor is close to going live, it is powered by IBM Cloud to provide global access.
The advisor uses AI visualization and machine learning to analyze seismic facies, characterize subsurface physical properties and assess exploration risk. The accuracy of these analyses directly relates to the knowledge captured by the system.
Deep knowledge adds value to exploration data
The development team trained the advisor to understand previous interpretations of seismic data as well as the collective knowledge and analytical workflows of our geoscientists. Thus, when analyzing exploration data, the advisor adds value to it by applying deep knowledge.
Because our whole team contributed knowledge, the system eliminates the subjectivity of individual analysts. This reduces uncertainty by standardizing analytical processes across all of Galp’s exploration projects, teams and geographies. And the advisor will become smarter over time as it learns from human interactions and ingests more data.
The goal of AI is not to replace humans but to complement them. Galp and IBM achieved this ideal by giving geoscientists a new way of working. Assisted by AI, the advisor unlocks the potential of exploration data to help analysts make better decisions—and, ultimately, to optimize resources and reduce risk.