CSV Weather Data

Source:: buem/weather/from_csv.py

For single-building runs, BuEM loads pre-extracted weather data from a CSV file rather than processing the full COSMO-REA6 grid.

CsvWeatherData

Reads CSV files with hourly columns:

T — air temperature [°C]
GHI — global horizontal irradiance [W/m²]
DNI — direct normal irradiance [W/m²]
DHI — diffuse horizontal irradiance [W/m²]

DNI Reconstruction

NWP models store DNI as \((GHI - DHI) / \cos(\theta_z)\). Near the horizon \(\cos(\theta_z) \to 0\), producing unphysical spikes (> 4 000 W/m² observed).

reconstruct_dni_from_ghi() applies pvlib’s DISC decomposition (Iqbal 1983) to re-derive DNI from GHI. The result is capped at extraterrestrial irradiance (1 316–1 413 W/m² seasonal). DHI is then back-computed as \(DHI = GHI - DNI \cos(\theta_z)\).

Caching

Processed data is stored as a .feather file next to the source CSV, avoiding the 2–3 s pvlib computation on repeated imports.

Default Weather File

BuEM ships with COSMO_Year__ix_390_650.csv — a representative weather year extracted from the COSMO-REA6 reanalysis for Loenen, Netherlands.