PV power forecasting using different Artificial Neural Networks strategies

Ines Sansa , Zina Boussaada (1, 2, 3) , Najiba Mrabet Bellaaj

1 UPV / EHU - University of the Basque Country = Euskal Herriko Unibertsitatea
2 ENIT - Ecole Nationale d'Ingénieurs de Tunis
3 ESTIA INSTITUTE OF TECHNOLOGY

The integration of photovoltaic (PV), intermittent and uncontrollable power, into the electrical grid has become one of the major challenges for power system operators. Therefore the PV power forecasting can be beneficial in system planning and balancing energies. In this paper the PV power forecasting of a real generator [1] is presented. Different Artificial Neural Networks (ANN) strategies are used to forecast the PV power from meteorological variables, the radiation and the temperature. Simulation results correspond to each ANN strategy are presented, discussed and compared. The Nonlinear Auto Regressive models with eXogenous input (NARX model) is the dynamic ANN chosen to use in this work. Its performances have proved in the different time frame PV power forecasting. The impact of type season's on PV power forecasting performances is also presented in the second part of this paper.

Mots clés: PV power forecasting; ANN; NARX model; model identification

Domaines: Sciences de l'ingénieur [physics]

Ines Sansa, Zina Boussaada, Najiba Mrabet Bellaaj. PV power forecasting using different Artificial Neural Networks strategies. 2014 International Conference on Green Energy, Mar 2014, Sfax, Tunisia. pp.54-59, ⟨10.1109/ICGE.2014.6835397⟩. ⟨hal-03036647⟩