Saved in:
Bibliographic Details
Main Authors: De Pellegrini, Vittoria, Alkhalifah, Tariq
Format: Preprint
Published: 2026
Subjects:
Online Access:https://arxiv.org/abs/2601.13190
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912905772924928
author De Pellegrini, Vittoria
Alkhalifah, Tariq
author_facet De Pellegrini, Vittoria
Alkhalifah, Tariq
contents Modeling and forecasting subsurface multiphase fluid flow fields underpin applications ranging from geological CO2 sequestration (GCS) operations to geothermal production. This is essential for ensuring both operational performance and long-term safety. While high fidelity multiphase simulators are widely used for this purpose, they become prohibitively expensive once many forward runs are required for inversion purposes and to quantify uncertainty. To tackle this challenge, we propose LAViG-FLOW, a latent autoregressive video generation diffusion framework that explicitly learns the coupled evolution of saturation and pressure fields. Each state variable is compressed by a dedicated 2D autoencoder, and a Video Diffusion Transformer (VDiT) models their coupled distribution across time. We first train the model on a given time horizon to learn their coupled relationship and then fine-tune it autoregressively so it can extrapolate beyond the observed time window. Evaluated on an open-source CO2 sequestration dataset, LAViG-FLOW generates saturation and pressure fields that stay consistent across time while running two orders of magnitude faster than traditional numerical solvers.
format Preprint
id arxiv_https___arxiv_org_abs_2601_13190
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle LAViG-FLOW: Latent Autoregressive Video Generation for Fluid Flow Simulations
De Pellegrini, Vittoria
Alkhalifah, Tariq
Machine Learning
Fluid Dynamics
Modeling and forecasting subsurface multiphase fluid flow fields underpin applications ranging from geological CO2 sequestration (GCS) operations to geothermal production. This is essential for ensuring both operational performance and long-term safety. While high fidelity multiphase simulators are widely used for this purpose, they become prohibitively expensive once many forward runs are required for inversion purposes and to quantify uncertainty. To tackle this challenge, we propose LAViG-FLOW, a latent autoregressive video generation diffusion framework that explicitly learns the coupled evolution of saturation and pressure fields. Each state variable is compressed by a dedicated 2D autoencoder, and a Video Diffusion Transformer (VDiT) models their coupled distribution across time. We first train the model on a given time horizon to learn their coupled relationship and then fine-tune it autoregressively so it can extrapolate beyond the observed time window. Evaluated on an open-source CO2 sequestration dataset, LAViG-FLOW generates saturation and pressure fields that stay consistent across time while running two orders of magnitude faster than traditional numerical solvers.
title LAViG-FLOW: Latent Autoregressive Video Generation for Fluid Flow Simulations
topic Machine Learning
Fluid Dynamics
url https://arxiv.org/abs/2601.13190