pygeomodeling 0.3.1

Advanced Gaussian Process Regression and Kriging toolkit for SPE9 reservoir modeling

PyGeomodeling

Advanced Gaussian Process Regression and Kriging toolkit for SPE9 reservoir modeling. This toolkit provides a comprehensive framework for spatial modeling of reservoir properties using both traditional Gaussian Process models and cutting-edge Deep Gaussian Process architectures.

Key Features

Core Capabilities

• GRDECL Parser: Load and parse Eclipse GRDECL files with automatic property extraction
• Unified Toolkit: Single interface supporting both scikit-learn and GPyTorch workflows
• Advanced Models: Traditional GP (RBF, Matérn, Combined kernels) and Deep GP with neural network features
• Rich Visualization: Comprehensive plotting utilities for model comparison and spatial analysis

Features (v0.3.0)

• Error Handling: Comprehensive validation with helpful error messages
• Model Serialization: Save/load models with versioning and metadata
• Spatial Cross-Validation: Proper validation for geostatistical data
• Parallel Processing: 3-4x speedup with multi-core training
• Hyperparameter Tuning: Optuna integration for optimization
• Variogram Analysis: Semi-variogram modeling for spatial correlation (NEW!)
• Tutorial Notebooks: Interactive learning with Jupyter
• Research-Ready: Built for reproducible scientific research

Installation

Basic Installation (Traditional GP Models)

pip install pygeomodeling

With Advanced Deep GP Features

pip install pygeomodeling[advanced]

Complete Installation (All Features)

pip install pygeomodeling[all]

Development Installation

git clone https://github.com/kylejones200/pygeomodeling.git
cd pygeomodeling
./setup_dev.sh  # Automated setup
# or manually: pip install -e ".[dev,docs,all]"

Quick Start

Basic Usage

from pygeomodeling import GRDECLParser, UnifiedSPE9Toolkit

# Load SPE9 dataset
parser = GRDECLParser('path/to/SPE9.GRDECL')
data = parser.load_data()

# Create toolkit and train model
toolkit = UnifiedSPE9Toolkit()
toolkit.load_spe9_data(data)
X_train, X_test, y_train, y_test = toolkit.create_train_test_split()

# Train traditional GP model
model = toolkit.create_sklearn_model('gpr', kernel_type='combined')
toolkit.train_sklearn_model(model, 'combined_gpr')
results = toolkit.evaluate_model('combined_gpr', X_test, y_test)

print(f"R² Score: {results.r2:.4f}")

Deep GP Experiment

from pygeomodeling import DeepGPExperiment

# Run comprehensive comparison
experiment = DeepGPExperiment()
results = experiment.run_comparison_experiment()

# Results automatically saved to deep_gp_comparison.png
print(f"Best model: {max(results.keys(), key=lambda x: results[x]['metrics']['r2_score'])}")

What's Included

Core Modules

• grdecl_parser.py: Eclipse GRDECL file parser with SPE9 support
• unified_toolkit.py: Main toolkit with sklearn and GPyTorch backends
• model_gp.py: Advanced GPyTorch models (Standard GP, Deep GP)
• plot.py: Comprehensive visualization utilities
• deep_gp_experiment.py: Ready-to-run comparison experiments

Example Scripts

• main.py: Traditional geomodeling workflow
• package.py: Packaging and distribution helper

Scientific Applications

This toolkit has been designed for:

• Reservoir Modeling: Spatial interpolation of permeability and porosity
• Uncertainty Quantification: Gaussian Process uncertainty estimates
• Model Comparison: Traditional vs Deep GP performance analysis
• Geostatistical Research: Advanced spatial modeling techniques

Model Performance

Based on SPE9 reservoir dataset analysis:

• Traditional GP (Combined Kernel): R² = 0.277 (Best)
• Deep GP (Small): R² = 0.189
• Training Time: ~1.3-1.8 seconds per model

Traditional Gaussian Process models with combined RBF+Matérn kernels outperform Deep GP models for SPE9 spatial patterns, demonstrating the effectiveness of classical geostatistical approaches for this dataset.

Dependencies

Core Dependencies (Always Installed)

• numpy >= 1.24.0
• pandas >= 1.5.0
• scikit-learn >= 1.3.0
• matplotlib >= 3.7.0
• pykrige >= 1.6.0

Optional Dependencies

• Advanced: torch, gpytorch, botorch, optuna (for Deep GP models)
• Geospatial: rasterio, geopandas, shapely (for advanced spatial analysis)
• Performance: numba (for JIT compilation)
• Visualization: matplotlib, signalplot (for publication-quality plots)

Documentation

Tutorial notebooks are available in the examples/ directory.

Contributing

Contributions are welcome! Please read our Contributing Guide for details on our code of conduct and the process for submitting pull requests.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Support

For questions, issues, or contributions:

• Issues: GitHub Issues
• Email: [e-mail]
• Documentation: Project Wiki

Citation

If you use this toolkit in your research, please cite:

@software{jones2025spe9geomodeling,
  title={SPE9 Geomodeling Toolkit: Advanced Gaussian Process Regression for Reservoir Modeling},
  author={Jones, Kyle T.},
  year={2025},
  url={https://github.com/kylejones200/pygeomodeling}
}

Outputs

• PERMX_GPR.GRDECL: predicted permeability in GRDECL format
• SIGMA_GPR.GRDECL: standard deviation (uncertainty) of predictions
• gpr_prediction_slices.png: side-by-side slice visualization