文档介绍：Solar aspects of papers
Use of artificial intelligence and an ad-hoc time series preprocessing
for solar insolation forecasting over a mountainous island
CYRIL VOYANT1,2, MARC MUSELLI2, CHRISTOPHE PAOLI2,
MARIE LAURE NIVET2 and PHILIPPE POGGI2
1Hospital of io, Radiotherapy Unit, 20177 io
2University of RS UMR SPE 6134
Vignola, Rte des Sanguinaires, 20000 io | Campus Grimaldi, 20250 Corte
FRANCE
christophe.******@univ- -
Abstract: - Artificial intelligence techniques are more and more used as alternative approaches as prediction conventional techniques or ponents of integrated systems. In this paper, we present a methodology for the prediction of daily and hourly global solar irradiation on a horizontal surface located in the mountainous island of Corsica. Our aim is to estimate and to manage photovoltaic power plants production connected to an work. We use an Artificial works (ANN) with a pre-treatment of the time series in view to predict solar radiation at different time horizons: 1 hour, 3 hour and 24 hour. First results are promising with nRMSE < 22% and RMSE < 129 Wh/m² for a prediction at 1h. On daily time-step, ANN present predicted profiles similar or even better than autoregressive models.
Key-Words: - Artificial works, Time series, Solar radiation, Forecasting, Stationarization, Complex terrain.
1 Introduction
T