Session

11.09.2017     16:30–18:30

Title:
MS20
Machine learning applications in subsurface reservoir modeling
Type:
Subsurface models (cryoshpere, hydrosphere, lithosphere, pedoshpere)
Room:

Machine learning (ML) have contributed significantly to recent advances in image and signal processing, pattern recognition, recommendation systems, natural language processing and machine translation. Most of these machine learning techniques, could be adapted for a wide range of applications in reservoir modeling. This Mini-symposium covers recent applications of machine learning algorithms for multi-scale modeling, reduced order modeling and uncertainty quantification (UQ) of subsurface reservoirs. Targeted applications includes: (1) Machine Learning assisted Uncertainty Quantification (2) ML accelerated statistical model calibration against multiple data sources (production, seismic, outcrops, experts) (4) Quantitative risk assessment using data-driven approaches. Also of relevance are Bayesian approaches, compressed sensing and sparse reconstruction methods, reduced-order parameterization, physical model cross-validation techniques, and response surface proxies.

11.09.2017
16:30–18:30

Title:
MS20
Machine learning applications in subsurface reservoir modeling
Type:
Subsurface models (cryoshpere, hydrosphere, lithosphere, pedoshpere)
Room:

Machine learning (ML) have contributed significantly to recent advances in image and signal processing, pattern recognition, recommendation systems, natural language processing and machine translation. Most of these machine learning techniques, could be adapted for a wide range of applications in reservoir modeling. This Mini-symposium covers recent applications of machine learning algorithms for multi-scale modeling, reduced order modeling and uncertainty quantification (UQ) of subsurface reservoirs. Targeted applications includes: (1) Machine Learning assisted Uncertainty Quantification (2) ML accelerated statistical model calibration against multiple data sources (production, seismic, outcrops, experts) (4) Quantitative risk assessment using data-driven approaches. Also of relevance are Bayesian approaches, compressed sensing and sparse reconstruction methods, reduced-order parameterization, physical model cross-validation techniques, and response surface proxies.


16:30–16:50
O101 A machine learning approach for uncertainty quantification using the Multiscale Finite Volume method
Ahmed H. Elsheikh (Edinburgh/GB)


16:50–17:10
O102 Efficient big data assimilation through sparse representation
Xiaodong Luo (Bergen/NO)


17:10–17:30
O103 A Data-Scalable Randomized Misfit Approach for Solving Large-Scale PDE-Constrained Inverse Problems
Tan Bui (Austin/US)


17:30–17:50
O104 Parametrization of Geological Models with Generative Advesarial Networks
Shing C. Chan Chang (Edinburgh/GB)


17:50–18:10
O105 Redistribution of steam injection in heavy oil reservoir management to improve EOR economics, powered by a unique integration of reservoir physics and machine learning
Pallav Sarma (Danville/US)


18:10–18:30
O106 DR-RNN: A deep residual recurrent neural network for model reduction
Nagoor K. Jabarullah Khan (Edinburgh/GB)