DeepCast AI at URTeC 2019 in Denver

July 19, 2019

During three exhibition days at URTeC 2019 at Denver, DeepCast shared its suite of Physics + AI solutions oriented to streamline field development decisions with several potential customers and partners. The booth was constantly busy and served for hosting the emergence of new business opportunities and synergies.

DeepCast URTeC 2019

DeepCast's booth

DeepCast differential technologies relies on a series of automated technologies including forecasting, RTA, uncertainty quantification, history matching, flow/geomechanics-informed AI predictions and optimization on unconventional assets. Thanks to the combination of AI with domain expertise in the business, DeepCast is able to unlock workflows as fast as 100x compared to traditional methodologies. Tangible savings in production and increase of ROI are realizable as more data is effectively used to generate integrated and much smarter solutions.

DeepCast’s plans to enrich its platform with more innovative and overarching solutions to the Oil and Gas business. Stay tune and see you at URTeC 2020, Austin TX!

Inferring Production in Undrilled Vaca Muerta locations with AI

DeepCast presented its paper “Improving Field Development Decisions in the Vaca Muerta Shale Formation by Efficient Integration of Data, AI, and Physics” at the session Machine Learning, AI, and Big Data (part V), URTeC 2019. This paper was written in collaboration with OpenSim and Primera Resources. The paper shows how a physics-based and AI-driven methodologies can both predict similar production trends on the same undrilled locations (used as blind-tests). The presentation was highly attended and generated interest with a few operators who approached the presenter, Hector Klie, after the end of the session.

DeepCast URTeC 2019

Figure 1. Production prediction of two undrilled pads using a full-fledged simulation and a AI-based methodology. Both results are right in the target for 3 years of production

The Figure 1 above shows that both physics-based and AI-based produced similar predictive results at the level of the pad (each pad had 3 wells). The predicted production corresponds to wells set apart as blind test and assumed to be undrilled. While the physics-based workflow required thousands of simulations to produce P50 type curve based on the production of neighboring wells, the AI workflow relied on digitized properties maps, automatic forecasting of approximately 800 wells, engineered feature extraction and mapping into production features. Once set, the AI workflow was capable of producing the prediction in matter of a few minutes. The physics-based workflow entailed about 5 days of work.

As an additional highlight, the work delivers meaningful business results with minimum public data and not access to proprietary data at all. This means that imposing adequate physical constraints and acute engineering judgement in combination with advanced machine learning technologies can potentially reduce stringent data requirements. Nevertheless, the whole study can be further improved with the aid of more diverse and higher resolution data that can potentially unravel a richer set of production indicators. These indicators are also instrumental for discovering an extended vision of the physics governing the fluid/fracture process in the Vaca Muerta formation and transferring learnings from different pads and formations. The presented paper complements the work “Leveraging US Unconventional Data Analytics Learnings in Vaca Muerta Shale Formation” previously presented in EUROPEC 2019, by some of the coauthors.

Related media content

deep cast article
Hector Klie chairs the SPE Reservoir Simulation Conference 2019
April 9, 2019
deep cast article brings AI to the SPE Workshop: Enhancing Production and Recovery from Mature Fields - The Mexican Perspective
February 21, 2019
deep cast article
DeepCast Shares Its Digital Transformation Perspective With the Mexican Petroleum Institute
February 18, 2019
+1 (833) 500-3282
1334 Brittmoore Rd
Houston, TX 77043
Copyright Ⓒ 2019 DeepCast, LLC