Integration of Physical Models and Machine Learning for photovoltaics: the publication on IEEE

Integration of Physical Models and Machine Learning for photovoltaics: the publication on IEEE

We announce the publication of the paper ” Experiments and Comparison of Digital Twinning of Photovoltaic Panels by Machine Learning Models and A Cyber-Physical Model in Modelica” in the scientific journal “IEEE Transactions on Industrial Informatics“.

The Dofware team, in collaboration with UNITO, has implemented as part of the HOME project a model capable of reproducing the behavior and productivity of photovoltaic panels.

The approach is the result of the cooperation between physical and Machine Learning models.

Modelica developed a digital twin of the photovoltaic panel focused on the reproduction of its physical properties, while with the application of Machine Learning techniques the data collected in the field were exploited to generate an alternative digital twin. The best result was obtained by integrating the two approaches, arriving at very accurate and reliable predictions.

This solution opens the way to new ways of development, in which physical-numerical and machine-learning models can quickly and effectively contribute to improvements that have so far been difficult to predict, not only in the photovoltaic field.

CONTACT US

For information and download:

https://ieeexplore.ieee.org/abstract/document/9525233