During the 2019 Data Science Conference, held on October 14-15, 2019, Bicheng Yan and Hector Klie presented the poster: “Automated Field Development Recommendation with Integration of Physics and AI”.
In this work, DeepCast describes the automation of workflows to deliver well-informed field development recommendations. The proposed technology allows engineers to perform optimization of different physical processes in a flexible manner, and ultimately enables them to play what-if scenarios between decision/uncertainty parameters and objective functions by advanced machine learning/AI models in an extremely efficient fashion. Recommendations involve automated data consolidation, physics-informed AI predictions, reservoir history matching, unsupervised uncertainty quantification via clustering and self-learning stochastic gradient optimizations.
During a highly visited poster session, DeepCast had the opportunity to show a new paradigm for streamlining field decisions associated with optimal completion/well design and drawdown under different attitudes towards risk.