UC2: Short-Term PV Forecasting – Results

The key operational KPI for UC2 is the wRMSE for PV production forecasts, which measures prediction accuracy relative to capacity. The results clearly demonstrate the step-change improvement delivered by incorporating EO data.

Table: Model comparison and performance in UC2

Model Configuration	Input Data Used	wRMSE Performance
Simple Baseline	Historical Patterns Only	22.1%
NWP Enhanced	Baseline + Numerical Weather Prediction (NWP)	16.1%
EO Enhanced	Baseline + EUMETSAT EO Data	9.7%
Best Configuration	EO + NWP Fusion	8.9%

The Best Configuration, fusing EO and NWP data, achieved a wRMSE of 8.9%. This is a significant improvement compared to the NWP-only model (16.1%) and the Simple Baseline (22.1%). 
These results confirm the initial hypothesis: EO data is essential to achieve the targeted operational utility, particularly in the short-term window where cloud movements drive PV ramps.