Petrinex Production Forecasting

ARPS decline curve analysis and production forecasting for oil and gas wells with comprehensive data validation.

Features

• ARPS Decline Curves: Automatic fitting of exponential, hyperbolic, and harmonic decline curves
• Smart Data Processing: Peak detection, outlier removal, and data quality filters
• High-Quality Forecasts: Average R² of 0.69 with 70% of wells achieving good fits (R² ≥ 0.6)
• Data Validation Framework: Bronze-to-silver data quality validation with change tracking
• Interactive Visualization: Streamlit app for exploring forecasts
• Audit & Compliance: JSON audit reports for data lineage and quality monitoring

Quick Start

1. Install Dependencies

uv sync

2. Validate Data Quality

uv run python run_validation.py

3. Run Forecasting

from petrinex.forecast import prepare_well_data, forecast_multiple_wells

# Load and prepare data
df = pd.read_parquet('fixtures/ngl_vol_bronze_cvx.parquet')
processed = prepare_well_data(df, min_months=6)

# Generate forecasts
forecasts = forecast_multiple_wells(processed, forecast_months=24, production_column='GasProduction')

User Interface

We have created a Streamlit app to allow users to explore the forecasts.

uv run streamlit run streamlit_app.py

Databricks Deployment

For deploying to Databricks:

databricks bundle validate
databricks bundle deploy --target dev
databricks bundle run

Data Validation Framework

The project includes a comprehensive data quality framework that:

• Bronze-to-Silver Validation: Automatically validates incoming data against silver tables
• Change Detection: Efficiently tracks data changes using row hashes
• Data Quality Checks: Validates data types, ranges, and handles NA values
• Audit Reports: Generates JSON audit trails for compliance and monitoring

Data Quality Pipeline

1. Bronze Tables: Raw Petrinex production data (ngl_vol_bronze_cvx.parquet, conv_vol_bronze_cvx.parquet)
2. Validation: Data type checking, range validation, NA handling, schema compliance
3. Silver Tables: Validated bronze data with metadata (hash, timestamps) for change tracking
4. Audit Reports: JSON files tracking data quality metrics and changes over time

All files are stored in fixtures/ directory for easy access and future Spark integration.

Validation Notebook

Run the complete validation pipeline using the Jupyter notebook:

uv run jupyter lab src/03_validate.ipynb

Or use the standalone script:

uv run python run_validation.py

Testing

Run tests with:

uv run pytest

Notes

• imperial unit conversions properly handled in the presentation layer (Streamlit app) rather than cluttering the core business logic.
• Logging should only happen within the codebase (src/petrinex) and not the notebooks.
• Works with volumes only, not using BDFS