Rational design using machine learning of luminescent materials for sunlight harvesting and energy conversion

• Autor(a) principal: Pereira, Bárbara Carvalho
• Data de Publicação: 2022
• Tipo de documento: Dissertação
• Idioma: eng
• Título da fonte: Repositório Científico de Acesso Aberto de Portugal (Repositórios Cientìficos)
• Texto Completo: http://hdl.handle.net/10773/38528
• topic: Solar; Renewable energy; Photovoltaic; Up-conversion; Quantum yield; Optical efficiency; Luminescent materials; Luminescent solar concentrators; Clustering; Machine learning

This dissertation aimed to study and develop the implementation of unsupervised Machine Learning algorithms and methods for application to luminescent materials for solar energy conversion. The Clustering algorithm is fast and reliable and it is demonstrated in this work its ease of application to databases. A bibliographic collection was carried out using the Web of Science platform of luminescent materials and their respective optical efficiency characteristics (Quantum Yield, Optical Efficiency, Stokes Shift and Power Conversion Efficiency). The algorithm was applied to the collected data in order to form clusters that allowed identifying and grouping the similarities between the defined figures of merit. As a result, it was viable to estimate the values of the figures of merit associated with each material by calculating the average value of the nearest neighbors in the same cluster for the same figure of merit. It was also possible to demonstrate the feasibility of the method to select materials that suit the needs of the cases under study, highlighting the luminescent solar concentrator incorporated with CuInS₂/ZnS reported in the literature with a quantum yield of 91%, power conversion efficiency of 2.94% and an absorption spectrum from ultraviolet to 830nm and emission between 620 to 1240nm. This work indicates the potential of applying machine learning models to materials science problems, although there are some limitations associated with the process, namely the fact that there are different ways of calculating the same parameter in the literature.