Research Overview

My research focuses on numerical modeling for subsurface energy systems, with underground hydrogen storage (UHS) as the central application. The work links thermodynamics, compositional multiphase flow, microbial reactive transport, and scientific machine learning across scales from phase-equilibrium calculations to reservoir-scale simulation.

A recurring theme is the trade-off between physical fidelity and computational efficiency: which numerical formulations remain accurate and robust under different coupling strengths, reaction time scales, and reservoir operating regimes?

Core Research Themes

From characterization to coupled simulation

Coupled Reactive-Compositional Reservoir Simulation

Development of reservoir-scale models coupling two-phase multicomponent flow with microbial reactive transport for UHS.

Multiphase flow Microbial reactions Reservoir simulation

Numerical Coupling Strategies

Comparison of fully implicit and sequential/operator-splitting formulations in terms of accuracy, robustness, computational cost, and regime dependence.

FIM SNI Operator splitting

Thermodynamics & Flash Acceleration

Hybrid physics-based and deep-learning-assisted methods for phase stability analysis and initialization of equilibrium calculations.

Phase equilibrium Flash calculation Deep learning

Reservoir Optimization

Data-driven surrogate modeling and optimization for cyclic UHS operation in depleted gas reservoirs.

Surrogate modeling Optimization Operations

Formation Evaluation & Petrophysics

Field-scale interpretation experience that informs reservoir characterization, model construction, and uncertainty-aware decision making.

Well logging Petrophysics Characterization
Current Numerical Focus

Fully implicit vs sequential/operator-splitting coupling

The current numerical-methods work compares fully implicit (FIM) and sequential non-iterative/operator-splitting (SNI) strategies for compositional flow coupled with microbial reactive transport. The comparison is organized around four questions:

  • Accuracy — How strongly does splitting error affect hydrogen loss and fluid-composition predictions?
  • Robustness — Under which reaction and transport regimes does each strategy remain stable?
  • Computational cost — When does reduced coupling complexity compensate for smaller time steps or additional splitting error?
  • Regime dependence — How do conclusions change with coupling strength, reaction stiffness, flow rate, and time scale?

Linear-solver and preconditioner improvements are treated as numerical infrastructure unless they materially alter this coupling-strategy comparison.

Related Publications