Weather, weather years and a typical year

Weather drives most of what the model calculates: how much heat the house loses, how much sun it gains, how much PV (solar panel) power it produces, and how hard a heat pump works. Every run uses one hourly weather file for one location. This article explains how that file is chosen and what it represents.

The location

The weather is fetched for your site. You give a country and a postal code in the house step, and the wizard resolves them to coordinates (latitude and longitude) and an altitude. The coordinates are rounded to two decimals, about a kilometre, because the weather sources are coarser than that anyway. The wizard prints the place name it found, so you can check it before you continue.

Two ways to answer

The weather step offers a typical year or a specific past year. They answer different questions.

A typical year tells you how the house behaves under normal conditions. It is assembled from the most ordinary months found in many years of records, so it never actually happened, which is the point: no single heatwave or mild winter skews the answer. It suits comparing envelope (walls, roof, floor and windows) or equipment options against a stable reference.

A specific year tells you what would have happened in that real year, with its own cold snaps and hot spells. It suits checking how an option behaves when the weather is unkind.

Which years are offered

The wizard offers five years: 2025, 2024, 2023, 2022 and 2021. Offering every year in the record would mean a wall of tiles, and each year costs a fresh data fetch at a new location.

Each year carries a short label and one sentence on what its weather was like across the countries we cover. Running the same house under 2022 and 2021 shows how far the weather alone can move a result:

2022 is marked as the heatwave year. Across Europe it was the hottest summer on record, according to the Copernicus Climate Change Service .
2021 is marked as a wet summer. June was hot, then July and August were wet and grey, with the floods in Belgium and western Germany. Across Europe as a whole, summer 2021 was a record warm one, mostly because of the heat in the south.

2025 also had strong heatwaves: June 2025 was the hottest June on record in western Europe, again according to Copernicus . Weather differs from one country to the next, so the hottest and coldest of these years at your address are worth checking in your own results.

Where the data comes from

Real years come from Open-Meteo’s historical weather service , which is built on ERA5 , a reanalysis from ECMWF (the European Centre for Medium-Range Weather Forecasts). A reanalysis is a modern weather model re-run over the past and combined with the observations of the time from across the world, such as weather stations (ground-based instrument sites, such as airports) and satellites. It gives a complete, consistent hourly value for every square of a regular grid, including places with no instrument, so it is not the reading of one station. ERA5 has a grid spacing of 0.25 degrees, about 25 km. Solar radiation comes from satellite-derived data, meaning sunshine estimated from satellite images of the clouds, which Open-Meteo documents for Europe at a finer resolution than the reanalysis grid.

For each hour the file holds air temperature, dew point (the temperature at which the air’s moisture starts to condense, a measure of how humid it is), humidity, global, direct and diffuse solar radiation (all the sunlight reaching a flat surface, the part coming straight from the sun, and the part scattered by the sky and clouds), wind speed and direction, and precipitation. The file is converted to the EnergyPlus weather format, the standard file type that building simulation programs read, and handed to the simulation.

How the typical year is built

We build the typical year ourselves, at your coordinates, from the last 15 complete calendar years. The method follows the one behind the TMY3 data sets (Wilcox and Marion, 2008, NREL/TP-581-43156) . TMY stands for typical meteorological year, and TMY3 is a widely used US collection of them. For each calendar month, we score every candidate year’s version of that month against the long-term distribution of temperature, dew point, wind and solar radiation, meaning how often each level occurs across all the years. Solar radiation is weighted most heavily. The best-matching version of each of the twelve months is kept, and the twelve are joined into one year. The joins are smoothed over a few hours so a cold December does not step straight into a mild January.

We apply the scoring stage only. Full TMY3 adds a second filter on persistence, such as runs of warm or cold days. We omit it, so the result is a typical year in the statistical sense and not a certified TMY3 file.

A first run at a new address takes longer while the file is built. The file is then stored, and later runs at the same location reuse it.

The weather is recorded with each run

Each saved run records which weather file produced it: the source (real year or typical year), the year for real years, the location it was built for, and a checksum of the file, a fingerprint of its contents. Two runs with the same identifier used the same weather, and a matching checksum shows the files were identical. On the results page the recap shows this as ERA5 followed by the year, or as Typical year. If you compare two runs, check that they used the weather you meant them to. Where dynamic prices (hourly prices that follow the wholesale market) are available, a real-year run is priced with prices from the same year.

Limits of the model

A reanalysis is not a measurement. It does not capture local effects such as a sheltered valley, a hilltop or the heat of a town centre.
Wind in the file is an average over a grid square and tends to be lower than at an exposed airport. Wind matters because it carries heat away from the outside walls and roof (convection).
A typical year smooths out extremes. It will understate the worst summer week, so run a real year if you care about one.
Each run uses one file. To see how weather changes your result, run the same house under different years and compare.
The weather is from the past. Future warming is not included, and neither is a changing climate over the life of the equipment.