A prototype developed by IBM in Brazil with Galp, through its subsidiary Petrogal Brasil, could soon become a global product in IBM's oil and gas portfolio, Ulisses Mello, director of IBM Research in Brazil, told BNamericas.
The prototype became ready earlier this year as part of a three-year research project to create an AI-based advisory tool to help geoscientists enhance seismic interpretation of exploration and development activities.
It consists of AI-based visual comprehension to assist geoscientists in sifting through large 3D seismic data sets and quickly identifying geological structures that could contain oil and gas.
The goal, as IBM reported on the occasion, was to reduce risk when assessing new prospects and optimize the placement of wells to be drilled.
The tool leverages Galp's acquired know-how and previous seismic interpretations and experiences, namely those acquired in the prolific pre-salt Santos basin. It continuously improves its capabilities, learning from the interaction with users and the insertion of additional data.
Galp's geoscientists are testing the prototype and additional capabilities are under joint development.
"This is a unique project. It has already been tested in several fields of Galp in Brazil and the results are very promising."
Inside IBM's Oil&Gas vertical, the idea is to create a product for the industry called Watson Natural Resources Platform that would incorporate the prototype.
According to Mello, it is not yet decided when and how the platform will be sold to other companies, which would include Galp's competitors. In the contract signed with IBM, Galp has the preference right to first use.
Part of the development of the solution, explains the scientist, comes from the knowledge gained with the training of Galp's geologists and geophysicists in the company's basins. Basins from other companies would require a new approach.
"If we are talking of other basins, there has to be another training. However, we work with a concept known as learning transfer, in which we assess the degree of accuracy for the replication of this training. A 50% rate is not good; 70% or more is better. In the case of this solution, we are seeing that the accuracy rate is high," he said.
Mello's IBM laboratory in Brazil is also developing solutions in partnership with companies such as Shell (in the field of shale gas, for US application), and for segments such as mining (with GoldenCorp) and agritech, as well as Watson-based solutions for the financial segment.
"The gravitational center of IBM in oil and gas research is Brazil," said Mello, who leads a team with over 100 researchers in the laboratory.