Regresar

On post-processing day-ahead NWP forecasts using Kalman filtering

Abstract:

Kalman filtering is an important concept in engineering and statistics. In the field of solar forecasting, it is well known as a numerical weather prediction (NWP) post-processing technique. However, it appears that this acknowledged post-processing technique needs some revisit. Since Kalman filtering is a sequential procedure, i.e., actual measurement from t is required to filter the forecast made for time t+1, it changes the forecast horizon of NWP from day-ahead to hour-ahead. Hence, the previously claimed improvements over NWP forecasts are not interpretable. Two simple remedies are proposed, which address the forecast horizon problem, but the effectiveness of the remedies is thought to be minimal.