Mar 27 2019
A team of researchers from the Spanish National Research Council and Universidad Politécnica de Madrid has come up with a novel approach based on artificial intelligence methods that account for the difference in the atmosphere when fabricating the solar cells to generate more amounts of energy.
By applying a combination of an artificial intelligence and statistical method, which is referred to as clustering, a research team from the Institute of Micro and Nanotechnology at Spanish National Research Council (IMN-CSIC) and the Solar Energy Institute at Universidad Politécnica de Madrid (IES-UPM) has identified a viable method to predict the generation of photovoltaic solar energy. The researchers accomplished this by including all the changes specified in the solar spectrum in their calculations.
Reported in Nature Communications, the study helps in finding an ideal design of multi-junction solar cells for every location, all in just a few hours of calculations. The atmospheric conditions and the position of the sun change all through the day and also all through the seasons of the year, and hence the sunlight reaching the photovoltaic panels tend to have varied properties. The spectral content of light experiences the most relevant change. It includes the distribution of light’s colors. For instance, the light is redder in the afternoon and bluer at midday.
The upcoming solar panels will be multi-junction and integrate numerous materials to manipulate the light spectrum. Yet, the energy production of multi-junction panels is dependent on the sunlight’s color change, to some extent.
Due to this reason, multi-junction panels are developed to generate the highest amount of energy for a specific color of the sunlight, and hence, the variations created by the atmospheric conditions and the position of the sun lead to production losses. In an effort to lower these production losses, scientists have developed a unique panel that overcomes the issue of the sunlight colors. This panel enables an optimal generation of global energy. Conversely, this optimization is extremely complex because of the countless range of atmospheric conditions together with the different sun positions.
Performed by the Spanish research team, the study has demonstrated that with the help of machine learning methods, data sets with many numbers of solar spectra can be reduced to a few typical proxy spectra. These proxy spectra can be effectively used for predicting the yearly averaged efficiency as a function of the design of solar cells.
The initial concept was developed by Iván García Vara (IES-UPM) at the time of his stay in the National Renewable Energy Laboratory. A statistical technique was developed by him to perform this type of calculation. Subsequently, the clustering method was applied to the earlier technique by Jose María Ripalda Cobián y Jeronimo Buencuerpo Fariña (IMN-CSIC) to achieve an effective outcome.
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.
Dr Iván García Vara, Researcher, Universidad Politécnica de Madrid