A Neural Network Post-processing Approach to Improving NWP Solar Radiation Forecasts - Archive ouverte HAL Accéder directement au contenu
Article Dans Une Revue Energy Procedia Année : 2014

A Neural Network Post-processing Approach to Improving NWP Solar Radiation Forecasts

Résumé

In this work, we investigate the use of artificial neural networks (ANNs) as a post-processing technique in order to improve mesoscale WRF solar radiation outputs.More precisely, one day ahead (with a 1 h temporal resolution) global horizontal irradiance (GHI) forecasts calculated by the WRF model are bias-corrected through the use of an ANN.ANNs are data driven approaches capable of recognizing patterns in data. Training data obtained from a ground station are used to construct the ANN model in order to reduce the bias of the WRF forecasts. A bias error analysis allows the determination of the relevant ANN's inputs necessary to this bias correction. Among others, one can cite the solar zenith angle and the clear sky index.This specific model output statistics (MOS) technique is applied in the frame of a solar PV forecasting project that takes place in La Reunion Island, a French oversea territory located in the Indian Ocean. This insular feature makes solar forecasting very challenging, hence the need for advanced solar forecasting methods.
Fichier principal
Vignette du fichier
Energy Procedia_lauret.pdf (473.82 Ko) Télécharger le fichier
Origine : Fichiers éditeurs autorisés sur une archive ouverte
Loading...

Dates et versions

hal-01089738 , version 1 (02-12-2014)

Identifiants

Citer

Philippe Lauret, Hadja Maïmouna Diagne, Mathieu David. A Neural Network Post-processing Approach to Improving NWP Solar Radiation Forecasts. Energy Procedia, 2014, 57, pp.1044-1052. ⟨10.1016/j.egypro.2014.10.089⟩. ⟨hal-01089738⟩
341 Consultations
956 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More