Solar simulation
Background
For our simulations, we need to be able to predict solar generation in the future for a certain asset. These assets do not exist yet, so training a machine learning model is impossible without past data. For this reason, we employ a physical solar simulation model that takes in input the asset characteristics and the weather data to simulate the asset generation.
The asset characteristics include location, DC capacity, inverter capacity, panel tilt and azimuth, and characteristics of the panel, such as temperature coefficient. We then use weather data from our weather model to calculate how much the panels would produce.
Methodology
We use technology originally developed at the US Sandia National Laboratories. Our technology has been extensively tested and validated with real data.
Weather variables
We use the following weather variables for the physical simulation:
| Variable | Unit | Time resolution |
|---|---|---|
| Air temperature at 2 m | °C | 1 hour |
| Wind speed | m/s | 1 hour |
| Global horizontal irradiance (GHI) | W/m2 | 1 hour |
| Diffuse horizontal irradiance (DHI) | W/m2 | 1 hour |
| Direct normal irradiance (DNI) | W/m2 | 1 hour |
Far-shading effects
By default, solar simulations account for terrain-based shading effects. We use horizon profile data from PVGIS for all assets to estimate the resulting loss of solar radiation caused by distant objects such as hills and mountains