amine-bouraoui/sofrilot-energy-intelligence

End-to-end energy cost attribution and forecasting system for an industrial cold-storage facility in Tunisia — SARIMAX attribution model, Medallion ETL pipeline, automated monthly PDF report, and Power BI dashboard. Real-world part-time engagement. Python · Power BI · DAX · Parquet · Open-Meteo API # Sofrilot Energy Intelligence System > End-to-end energy cost attribution and automated reporting system for an industrial cold storage facility in Tunisia. ## Business Context Sofrilot is a cold storage and logistics company in Tunisia that rents positive and negative refrigeration rooms to other businesses. Negative rooms are used by ice companies as collection points, resulting in high door activity especially in summer. Solar panels were installed after opening, creating a natural before/after intervention point. ## Project Goal Build a model that attributes electricity costs to their drivers: - Solar panel production offset - Ambient temperature (seasonal demand) - Room door activity (ice company behavior) - Occupancy rate - Defrost cycle frequency ## Data Sources | Source | Description | Granularity | |--------|-------------|-------------| | XWeb / Dixell | Temperature sensor logs for all cold rooms | 5-minute intervals | | Electricity meter | kWh consumption for the whole facility | Hourly | | Solar inverter | kWh production from solar panels | Hourly | | Open-Meteo API | Historical weather data for Tunis coordinates | Hourly | | Internal records | Room occupancy log per client | Daily | ## Project Structure ``` sofrilot-energy-intelligence/ ├── data/ │ ├── bronze/ # Raw files as received │ ├── silver/ # Cleaned and typed data │ └── gold/ # Master table ready for modeling ├── notebooks/ │ ├── 01_understanding.ipynb │ ├── 02_cleaning_pipeline.ipynb │ ├── 03_feature_engineering.ipynb │ ├── 04_cost_attribution_model.ipynb │ ├── 05_forecasting.ipynb │ └── 06_automated_report.ipynb ├── pipeline/ │ └── generate_report.py ├── reports/ │ └── sample_report.pdf └── dashboard/ └── sofrilot_dashboard.pbix ``` ## Methodology 1. **Bronze → Silver → Gold** ETL pipeline for all data sources 2. **Feature engineering** — door activity proxy from temperature volatility, net electricity cost after solar offset 3. **SARIMAX cost …