Machine Learning for BIPV Production

  1. D. Granados-López
  2. D. Gonzalez-Peña
  3. A. García-Rodríguez
  4. S. García-Rodríguez
  5. M. García-Fuente
Actas:
38th European Photovoltaic Solar Energy Conference and Exhibition

ISBN: 3-936338-78-7

Año de publicación: 2021

Páginas: 1529 - 1531

Tipo: Aportación congreso

Resumen

The increase in energy consumption in homes has never been as intense as in the last decade. The international alarm has resulted in various plans and strategies. This research supports the implementation of Building Integrated Photovoltaics (BIPV) which produce energy at a lower cost to the environment than traditional fuel. Also, it generates energy where it is needed, so its implementation preserves the planet's health. This research proposes a model to predict vertical PV production from vertical solar irradiation and ambient temperature measures. The predictor is an Artificial Neural Network (ANN), trained using the Levenberg-Marquardt algorithm. The research tested and compared several architectures to obtain the most accurate ANN. The one chosen was an artificial neural network with a single hidden layer. Its entry has two neurons, ambient temperature and global vertical irradiation (RaGV). This research shows that ambient Temperature and RaGV are enough to estimate BIPV production, being RaGV the most relevant variable. The introduction as input of the model of the ambient temperature increased its performance. The obtained model has good statistical results and allows to estimate the PV production of commercial panels.