Optimizing the energy production of photovoltaic panels using artificial intelligence

(Nanowerk News) By using a statistical technique and artificial intelligence, which is known as clustering, researchers from Solar Energy Institute at Universidad Politécnica de Madrid (IES - UPM) and from Institute of Micro and Nanotechnology at Spanish National Research Council (IMN-CSIC) have found a practical way to include in their calculations all the changes given in the solar spectrum to predict the production of photovoltaic solar energy.
The study, published in Nature Communications ("Solar cell designs by maximizing energy production based on machine learning clustering of spectral variations"), allows us to find, in a few hours of calculating, the optimal design of multi-junction solar cells for each location.
Throughout the day and the seasons of the year, the sun position and the atmospheric conditions change and thus the sunlight that reaches the photovoltaic panels have different characteristics. The most relevant change occurs in the spectral content of light which consists in the distribution of colors of light. For example, the light is bluer at midday and redder in the afternoon.
The future solar panels will be multi-junction and combine various materials to take advantage of the light spectrum. However, the production of energy of multi-junction panels depends to some extent on the color change of the sunlight.
For this reason, these panels are manufactured to produce the maximum energy for a certain color of the sunlight, and thus the changes produced by the sun position and the atmospheric conditions cause production losses.
In order to reduce these losses, researchers have designed a panel with an optimal production of global energy that it solves the problem of the sunlight colors. However, due to the infinite variety of atmospheric conditions combined with the diverse sun positions, this optimization is very complex.
The work carried out by the Spanish researchers has shown that data sets with thousands of solar spectra can be reduced to a few characteristic proxy spectra using machine learning techniques, and successfully use these proxy spectra to predict the yearly averaged efficiency as a function of the solar cell design.
Iván García Vara (IES - UPM) came up with the initial idea during his stay in the National Renewable Energy Laboratory. He developed a statistical method to conduct this type of calculation ("Spectral binning for energy production calculations and multijunction solar cell design"). Later, Jose María Ripalda Cobián y Jeronimo Buencuerpo Fariña (IMN - CSIC) applied the clustering technique to the previous method to achieve a successful result. Iván García Vara points out “the final result of our work project was the optimization of the design of multi-junction solar panels using the yearly energy production as a criterion”.
Source: Universidad Politécnica de Madrid
Subscribe to a free copy of one of our daily
Nanowerk Newsletter Email Digests
with a compilation of all of the day's news.