Applying deep reinforcement learning to seismic interpretation was the focus of this technical presentation by Roderick Perez, Senior Expert Geoscientist, who shared research from his Master’s thesis in Data Science on using AI to support strategic drilling and geological exploration.
The presentation explored how reinforcement learning can be applied to seismic amplitude data to help identify potential drilling locations. Roderick began by explaining how geoscientists traditionally interpret seismic data before introducing an AI approach capable of learning from synthetic geological environments that replicate different subsurface conditions.
A significant part of the session focused on the development of those training environments. Roderick demonstrated how geological processes, including deformation, faulting, tilting and fluid migration, were modelled to generate realistic seismic scenarios. These synthetic models were then used to train a Deep Q Network, allowing the AI agent to learn how different geological configurations influence drilling decisions.
The presentation concluded with demonstrations of the trained model, showing how it evaluated seismic images, navigated different geological environments and progressively learned to identify commercially attractive drilling targets through repeated interaction with the simulated data.
Watch the full presentation to discover how deep reinforcement learning can be applied to seismic interpretation and strategic drilling using realistic geological models and AI-driven decision-making.
Speaker
Roderick Perez, Senior Expert, Geoscientist, Energy Technology
