Comparison of irradiance forecasts from operational NWP model and satellite‐based estimates over Fennoscandia

Abstract

AbstractAccurate irradiance forecasts are needed for the growing solar energy industry also in Northern Europe. We have compared irradiance forecasts from an operational numerical weather prediction (NWP) model, a satellite‐based model, and persistence models. We aim to determine whether operational NWP models are suitable for forecasting irradiance at the high latitudes, and how their accuracy compares to the satellite‐based model. We have included all members and the ensemble average of the MetCoOp ensemble prediction system (MEPS), the MetCoOp‐Nowcasting (MNWC) system, the satellite‐based Solis‐Heliosat model, and two persistence models. The comparison is made as a point comparison against in situ irradiance observations in Finland and Sweden, for intra‐day forecasts with hourly and 15‐min output and the full forecast of MEPS with hourly output. In addition, we show two energy market case studies. We find the operational NWP models to be very suitable for irradiance forecasting in the area, up to the full horizon of the forecasts. Solis‐Heliosat errors grow with lead time, while the NWP model errors are largest in the beginning, settling to smaller values after the first hours. Solis‐Heliosat has more accuracy for the first 2–3 h of the forecast, after which NWP models produce better forecasts. However, during morning periods Solis‐Heliosat is found to have limited accuracy, while conversely, MNWC performs better in the morning than in the afternoon. The energy market case study highlights the same results: NWP models do well with forecasting irradiance in Fennoscandia, but the optimal selection of forecast model depends on the required forecast horizon and time.