Bayesian neural network surrogates for Bayesian optimization of Carbon Capture and Storage operations

Sofianos Panagiotis Fotias

doi:10.1016/j.egyr.2026.109549

Bayesian neural network surrogates for Bayesian optimization of Carbon Capture and Storage operations

2026-07-22

0

OA

AI

S

Sofianos Panagiotis Fotias *

V

Vassilis Gaganis

DOI：10.1016/j.egyr.2026.109549

Original

Original request for help

Save

Abstract

En 中文

• Benchmarked GP and several uncertainty-aware neural surrogates for Bayesian optimization in CCS operations. • BO delayed net-sequestration decline by ∼10 years and increased stored CO2 from ∼1.9 to ∼2.3 Tscf. • Long-horizon scheduling ( ≈1000 controls) recovered ∼1.76 Tscf stored CO2 with five fewer producers. • NPV improved from ∼€1.15B to ∼€1.36B, reaching ∼90%–93% of the placement-optimum benchmark. • Constraint-induced control clipping and recycling-dominated solutions are key failure modes for CCS BO surrogates.

Keywords:

CCS

Bayesian optimization

Markov chain Monte Carlo

Variational inference

Gaussian process

Reservoir simulation

Bayesian neural network surrogates for Bayesian optimization of Carbon Capture and Storage operations

Abstract

Journal

Organization

Cited Papers

Citing Papers