Solar irradiation nowcasting by stochastic persistence: a new parsimonious, simple and efficient forecasting tool

Cyril Voyant (1) , Gilles Notton (1)

1 SPE - Sciences pour l'environnement

Simple, naïve, smart or clearness persistences are tools largely used as naïve predictors for the global solar irradiation forecasting. It is essential to compare the performances of sophisticated prediction approaches with that of a reference approach generally a naïve methods. In this paper, a new kind of naïve " nowcaster " is developed, a persistence model based on the stochastic aspect of measured solar energy signal denoted stochastic persistence and constructed without needing a large collection of historical data. Two versions are proposed: one based on an additive and one on a multiplicative scheme; a theoretical description and an experimental validation based on measurements realized in Ajaccio (France) and Tilos (Greece) are exposed. The results show that this approach is efficient, easy to implement and does not need historical data as the machine learning methods usually employed. This new solar irradiation predictor could become an interesting tool and become a new member of the solar forecasting family.

Mots clés: prediction; machine learning; forecasting; persistence; bias-variance

Domaines: Recherche d'information [cs.IR]; Intelligence artificielle [cs.AI]; Sciences de l'ingénieur [physics]

Cyril Voyant, Gilles Notton. Solar irradiation nowcasting by stochastic persistence: a new parsimonious, simple and efficient forecasting tool. Renewable and Sustainable Energy Reviews, 2018, 92, pp.343-352. ⟨hal-01783937⟩