Open-Meteo likely overestimates winter GHI in Northern Italy's Po Valley due to frequent fog, which reduces irradiance but may not be fully captured in its models.[1][2] PVlib ModelChain parameters may also need tuning for local Italian PV mixes, such as temperature coefficients from JRC Ispra data.[3] Seasonal calibration beyond a single factor is key, using better weather data like Solcast or ERA5.[4][5]
## Weather Data Issues
Po Valley fog in winter (late fall to early spring) creates persistent low-level temperature inversions and radiative cooling, severely cutting GHI—often ignored in coarser models like Open-Meteo.[1] Switch to Solcast for direct PV forecasts with ~10% better European accuracy via high-res ICON-EU (6.5km), or ERA5 reanalysis validated for Italy.[4][5] ENEA Solaritaly offers measured GHI/DNI/DHI for 243 Italian sites (2006-2022), ideal for baselines.[6]
## PVlib Calibration Tips
Use province-specific lat/lon/tilt from ProfileSolar (e.g., Pavia: fog-aware analysis).[7][2] Fit seasonal temperature models (e.g., PVSyst) and spectral losses with JRC coefficients from Ispra (near Po Valley).[8][3] For aggregate ~9.8 GW, apply post-processing: train ML regressor on ENTSO-E actuals vs. your ModelChain output, adding solar elevation corrections to fix winter bias.[9]
## Parameter Adjustments
| Aspect | Winter Fix | Summer Fix | Source |
|--------|------------|------------|--------|
| GHI Input | Scale by 0.65 or fog proxy (e.g., RH inversion) | Minimal change | [1] |
| Temp Coeff | Higher losses (use n-type bifacial TCs from JRC) | Standard | [10] |
| Soiling/Other | Increase to 5-10% (PM from Po Valley) | 2-5% | [11] |
## Alternatives & Next Steps
- **Weather**: Solcast API (direct PV kW, 30min res).[11]
- **Full Pipeline**: EMHASS-style forecast adjustment on ENTSO-E.[9]
- **Validate**: Compare vs. ENEA tables per province; aggregate >=1MW subset.[6]
Test ERA5 via PVlib's `get_clearsky` with measured baselines for quick wins.[12]
Citations :
Ai