Ocean Framework, C#, .NET, C++, CUDA
History matching is a critical step in reservoir engineering. It consists of calibrating numerical models so that they reproduce observed field data, such as production, 4D seismic, and well logs, reducing uncertainty in reservoir behavior forecasts. Petrobras required a tool that applied modern probabilistic methods based on the Kalman Filter while remaining fully integrated with Petrel, the central platform for its geosciences and reservoir engineering workflows.
Implementing data assimilation methods at industrial scale poses significant technical challenges: handling hundreds of reservoir model realizations per iteration, integrating multiple heterogeneous data sources (production, 4D seismic, well logs), running parallelized simulations on high-performance computing (HPC) clusters, and ensuring compatibility with different reservoir simulation engines. All of this while preserving a workflow native to Petrel, with no intermediate exports that would compromise data traceability and team productivity.
DeepSoft developed BR-Kalman, a native Petrel plug-in that implements ensemble-based history matching methods derived from the Kalman Filter, with emphasis on the ES-MDA (Ensemble Smoother with Multiple Data Assimilation) technique. Built on the Ocean Framework, the solution updates reservoir models by simultaneously assimilating production, 4D seismic, and well log data, generating multiple calibrated models that honor the observed history.
The architecture was designed to scale in enterprise environments: BR-Kalman integrates with high-performance computing (HPC) clusters for parallel execution of realizations, operates with the CMG family of reservoir simulators (IMEX, GEM, STARS), and is being adapted to the SLB DELFI ecosystem, enabling hybrid workflows between the local Petrel environment and the cloud.
Generation of multiple calibrated models that simultaneously honor different types of historical data, providing a robust statistical basis for uncertainty quantification.
The entire assimilation cycle, from parameterization to results analysis, runs inside Petrel, eliminating manual exports and preserving coherence with the project's geological and simulation models.
Parallel execution of ensembles on HPC clusters and support for CMG simulators, enabling history matching in complex reservoirs with large numbers of uncertain parameters.
In recognition of the results delivered, DeepSoft signed a 5-year Technological Order contract with Petrobras for the continued support, evolutionary maintenance, and new feature development of BR-Kalman, including its adaptation to the SLB DELFI ecosystem. The partnership consolidates the collaboration with CENPES (Petrobras Research, Development and Innovation Center) on a highly relevant project for the Oil & Gas industry.