Machine Learning (ML) in Reservoir Engineering

1. In the context of reservoir modeling, can we treat the ‘conceptual model’ as “actions”, where ML would treat each physical, chemical or biological process as a ‘logical term’? &

Can we represent the ‘mathematical model’ (partial differential equation describing the fluid flow through a petroleum reservoir) deduced from the fundamental ‘differential calculus’ to be treated as “agent’s knowledge”, where ML would treat each mathematical equation as a ‘logical formula’?

Feasible to characterize a petroleum reservoir using such ‘logic’ that would probably allow full introspection of reservoir beliefs at various scales?

2. Whether interpretation or explanation of ‘reservoir management’ decisions produced by ML learning models (Explainable AI / Clustering) would really help - in disclosing, any additional details - that was left unnoticed (or failed to capture) by the reservoir engineering management team?

Whether ‘post-modeled Explainable AI’ (explainable clustering based on unsupervised decision trees, where a threshold tree provides a clear interpretation of the resulting clustering; unlike the explanation provided by the relatively cheaper ‘small decision tree’) would focus on giving the reasoning behind the decisions/output made by ‘petroleum software packages’ (black box models)?

3. In the context of reservoir fluid properties, if only one fluid (water or oil or gas) is allowed to occupy the same location (given pore space) at the same time in a petroleum reservoir, how could we make use of the concept of Multi-Agent Meeting problem (MAM) in petroleum reservoir applications 'effectively'?

4. In the context of reservoir rock properties, can we efficiently make use of the concept of Multi-Directional Meet (MDM) towards estimating ‘reservoir permeability’?

Feasible to deduce ML algorithms that essentially progresses along the principal fluid flow directions following the least resistive (tortuous) pathways that actually represents the reservoir permeability (rather than deducing reservoir permeability over a REV)?

More Suresh Kumar Govindarajan's questions See All
Similar questions and discussions