Since the beginning of this century, the share of renewables in Europe's total power capacity has almost doubled, becoming the largest source of its electricity production. In 2015 alone, photovoltaic (PV) energy generation rose with a rate of more than 5%; nowadays, Germany, Italy, and Spain account together for almost 70% of total European PV generation. In this context, the so-called day-ahead electricity market represents a key trading platform, where prices and exchanged hourly quantities of energy are defined 24 h in advance. Thus, PV power forecasting in an open energy market can greatly benefit from machine learning techniques. In this study, the authors propose a general procedure to set up the main parameters of hybrid artificial neural networks (ANNs) in terms of the number of neurons, layout, and multiple trials. Numerical simulations on real PV plant data are performed, to assess the effectiveness of the proposed methodology on the basis of statistical indexes, and to optimize the forecasting network performance.

ANN sizing procedure for the day-ahead output power forecast of a PV plant

Grimaccia, Francesco;Leva, Sonia;Mussetta, Marco;Ogliari, Emanuele
2017-01-01

Abstract

Since the beginning of this century, the share of renewables in Europe's total power capacity has almost doubled, becoming the largest source of its electricity production. In 2015 alone, photovoltaic (PV) energy generation rose with a rate of more than 5%; nowadays, Germany, Italy, and Spain account together for almost 70% of total European PV generation. In this context, the so-called day-ahead electricity market represents a key trading platform, where prices and exchanged hourly quantities of energy are defined 24 h in advance. Thus, PV power forecasting in an open energy market can greatly benefit from machine learning techniques. In this study, the authors propose a general procedure to set up the main parameters of hybrid artificial neural networks (ANNs) in terms of the number of neurons, layout, and multiple trials. Numerical simulations on real PV plant data are performed, to assess the effectiveness of the proposed methodology on the basis of statistical indexes, and to optimize the forecasting network performance.
2017
Artificial neural network; Day-ahead forecast; Ensemble methods; Fluid Flow and Transfer Processes; Process Chemistry and Technology; Computer Science Applications1707 Computer Vision and Pattern Recognition; Engineering (all); Materials Science (all); Instrumentation
File in questo prodotto:
File Dimensione Formato  
ANN Sizing_applsci-07-00622.pdf

accesso aperto

: Publisher’s version
Dimensione 2.79 MB
Formato Adobe PDF
2.79 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11311/1041034
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 48
  • ???jsp.display-item.citation.isi??? 38
social impact