DeepCast Invited to Present Physics-Informed AI at 2019 AGU Fall Meeting

November 20, 2019

DeepCast sponsored, co-organized and presented at the SPE Workshop: Data-Driven and Physics-Based Models for Enhanced Reservoir Insights and Predictions, held in San Antonio, TX, on November 19-20.

This workshop represents the first of its kind to focus on key issues that may lead to the generation of interpretable and generalizable models to extend the current modeling, forecasting and optimization capabilities in the Oil and Gas industry. The event participation was sold-out one month in advance and counted on the participation of more than 100 attendees coming from 4 continents and more than 60 institutes, including major operators and service companies in Oil and Gas, several academic institutions, national labs and tech solution providers in the space of machine learning and artificial intelligence.

DeepCast, represented by Hector Klie, co-founder and CEO, co-chaired the session “Combining Data-Driven and Physics-Driven Models” and a wrap-up Panel Discussion dedicated to explore future opportunities that could be unlocked with forthcoming advances in algorithmic, software and hardware technologies.

In this workshop, Hector Klie presented DeepCast's modeling work entitled: “AI-Based Field Development and Optimization in Unconventional Reservoirs”.

This work entails a novel approach to generate physics-informed AI models that effectively combines physics-based and data-driven models with domain expertise. These models have the capability to accurately represent coupled dynamics involved in multi-well interactions, multiple phases and other spatiotemporal subsurface processes. The proposed new class of models offers highly desired features such as: (1) amenability to accommodate both field and simulation data; (2) flexibility to handle an arbitrary number of decision and uncertainty parameters; (3) resilience to overfitting; (4) simplicity; (5) interpretability; (6) speed-ups of the order of 1000x with respect to regular simulations and; (7) transferability to different field planning and optimization scenarios.

‍

‍

Fig 1. Hector Klie presenting at the SPE Workshop: Data-Driven and Physics-Based Models for Enhanced Reservoir Insights and Predictions. (Plenary view)

‍

Fig 2. Hector Klie presenting at the SPE Workshop: Data-Driven and Physics-Based Models for Enhanced Reservoir Insights and Predictions. (Close-up)