What It Does

This project forecasts European LNG (liquefied natural gas) demand at a city level using a multi-source dataset combining weather, energy usage, and financial disclosures.

It models heating degree days (HDD) to estimate heating-related demand and forecasts near-term LNG usage across European regions — useful for traders, utilities, and policy analysts.

How It's Built

• Data Sources:

  • Eurostat: country level energy consumption data
  • BrightSky: city level historical + forecast weather data
  • yr.no: city level historical + forecast weather data
  • SEC EDGAR: LNG import/export details from company filings

• Feature Engineering:

  • Heating Degree Days per city modelled using city level weather data
  • Lagged weather + usage patterns
  • Cross-region trade and demand interaction features

• Pipeline + Orchestration:

  • ETL built in Python using requests, pandas, and NumPy
  • Fully orchestrated in Dagster, with assets for each data source and forecast stage
  • Scheduled via Dagster cron integration for weekly forecast refreshes

• Modeling:

  • Gradient boosting model trained on 5+ years of weather and usage data
  • Outputs regional LNG demand estimates for the next 30–60 days

• Output:

  • Results are stored to local dashboards
  • Key forecasts optionally pushed to a Discord channel as alerts

Screenshot

Why It Matters

Europe's LNG demand is highly seasonal and sensitive to weather volatility. This tool helps:

• Anticipate regional heating demand spikes
• Adjust forecasts in response to early cold snaps or mild winters

The project is available on GitHub