Data Center Water Efficiency (WUE) Analysis

A datathon project that predicts optimal data center locations in the US based on Water Usage Efficiency (WUE = Annual Water Usage L / IT Energy kWh), using climate and grid-renewable data.

Approach

1. Regression model — OLS on African data center data: WUE ~ wetbulb temperature + wind speed + precipitation + renewable energy share
2. US climate data — Weather via Open-Meteo API for US climate divisions
3. Predictions — Apply the model to US climate divisions to score locations

Key Files

File	Description
datacenter_project.ipynb	Main notebook: WUE model + US predictions
wue_regression.py	Standalone OLS regression script
scrape_weather.py	Fetches climate data (wetbulb, wind, precip) from Open-Meteo
plot_climate_regions.py	Map of predicted WUE by US climate division
plot_world.py	Map of grid carbon intensity + renewable plants + WUE sites

Setup

pip install pandas geopandas matplotlib statsmodels openmeteo-requests requests-cache retry-requests cmocean

Data

• data/daily_african_df.csv — African data center climate + WUE (training data)
• data/climate_data.csv — US climate division aggregates (from scrape_weather.py)
• data/CONUS_CLIMATE_DIVISIONS/ — US climate division shapefiles
• geojson_files/ — Electricity zones, water efficiency points