Use real weather

What “real weather” means here

For dates within the next ~16 days, solarprod fetches the forecast from a numerical weather prediction model (Open-Meteo / GFS-style). For dates further in the future it falls back to a 5-year historical average for that location and time of year — a “typical year” climate baseline. For past dates it uses real reanalysis data, so a backtest against your actual production should match tightly.

The simulator translates cloud-cover percentages into irradiance via a clearness-index model, then runs the standard solar geometry calculation per panel group. Multi-day chained forecasts apply confidence-band uncertainty that gets wider further out — with decreasing confidence at later dates.

When to turn it off

Leave it on for everyday use — it’s the realistic answer. Turn it off only when you specifically want to:

• Compare panel layouts head-to-head without weather noise confusing the result — a clear-sky run shows the geometric ceiling for each layout.
• See your system’s “best-case” output to compare against an installer’s claim.
• Sanity-check the calculator itself — a clear-sky run should produce a smooth, symmetric curve.

Don’t use clear-sky as your day-to-day forecast — most populated parts of the world average roughly 50– 70 % of clear-sky output across the year (sunnier in arid and equatorial regions, lower in marine and high-latitude ones), with substantial day-to-day variance.