Multipliers Zonation for Reservoir History Matching

This repository is the companion codebase for our paper "Multipliers Zonation for Reservoir History Matching". It contains the full workflow we used in the study: optimization code, zonation strategies, deck variants, generated permeability files, logs, and figure outputs used for analysis and reporting.

What you can reproduce here

• Zonation-aware history matching with iterative refinement + L-BFGS multiplier optimisation
• A no-zonation baseline (per-cell multipliers)
• Train/test behaviour across prepared deck variants
• Outputs used in the analysis: permeability fields, rate/BHP curves, optimisation logs

Visual Overview

How the model is divided into zones

Rates and BHP comparison across zoning strategies

Installation

Supported Software Versions

• Julia: 1.12.0 (the checked-in Manifest.toml was generated with this version).
• OS: Linux/macOS/Windows with a working Julia installation.
• Graphics: for interactive 3D plotting (GLMakie), an OpenGL-capable environment is required.
• Headless environments: plotting calls may need to be disabled or procede to use CLMakie.

Install Steps

1. Clone the repository and enter it:

   git clone https://github.com/deepfield-team/History-matching.git
   cd History-matching

2. Instantiate dependencies:

   julia --project=. -e 'using Pkg; Pkg.instantiate(); Pkg.precompile()'

Getting Started

1. Run zonation-aware history matching (main workflow)

julia --project=. experiments/code_spe1_zone_LBFGS.jl

What this run does:

• Loads SPE1 reference data.
• Runs iterative refinement + L-BFGS multiplier optimization.
• Writes permeability include file(s) into models/inc/.
• Writes logs and rate curves into experiments/logs/.
• Displays plots (loss, gradients, permeability, rates).

2. Run no-zonation baseline

julia --project=. experiments/code_spe1_no_zone.jl

This optimizes per-cell multipliers directly and writes models/inc/spe1_no_zone_perm.inc.

3. Compare train/test behavior across deck variants

julia --project=. experiments/code_spe1_train_test_rates.jl

This script simulates prepared deck variants and plots train/test curves for selected QoIs.

4. Change refinement strategy

Edit experiments/code_spe1_zone_LBFGS.jl:

• Set ZONING_TECHNIQUE_KEY to one of: sign_uncons, sign_cons, medium-cons, medium_uncons, hierarchical clustering.

Regenerating include files

Running the optimization scripts regenerates include files automatically:

• Zonation workflow writes files like models/inc/spe1_zone_LBFGS_perm_<technique>.inc.
• No-zonation workflow writes models/inc/spe1_no_zone_perm.inc.

Repository Structure

.
├── src/
├── experiments/
├── models/
├── pics/
├── Project.toml
├── Manifest.toml
└── README.md

Data Files: Where to Get Them

• Base SPE1 deck: models/SPE1.DATA (used from GeoEnergyIO).
• Variant deck (sign constrained): models/SPE1_sign_cons.DATA
• Variant deck (sign unconstrained): models/SPE1_sign_uncons.DATA
• Variant deck (median constrained): models/SPE1_medium-cons.DATA
• Variant deck (median unconstrained): models/SPE1_medium_uncons.DATA
• Variant deck (hierarchical clustering): models/SPE1_hierarchical clustering.DATA
• Variant deck (no zoning): models/spe1_no_zone.DATA
• Permeability include files referenced by variant decks: models/inc/*.inc

experiments/ (entry-point scripts and outputs)

• experiments/code_spe1_zone_LBFGS.jl: main zonation-aware optimization pipeline.
• experiments/code_spe1_no_zone.jl: no-zonation baseline optimization.
• experiments/code_spe1_train_test_rates.jl: train/test comparison across prepared decks.