In the quest for more efficient and cost-effective solar energy solutions, researchers from the University of Cambridge, including Lucas Li, Jean-Baptiste Puel, Florence Carton, Dounya Barrit, and Jhony H. Giraldo, have developed a new machine learning model to predict the performance of perovskite solar cells. Their work, published in the journal Nature Communications, aims to accelerate the development of these promising next-generation photovoltaics.
Perovskite solar cells hold great potential due to their high efficiency and low production costs. However, their performance is influenced by complex interactions between their various layers, making the traditional trial-and-error approach to finding the best combinations of materials and architectures slow and expensive. To tackle this challenge, the researchers have introduced a model called Solar-GECO, which stands for Solar-Geometric-Aware Co-Attention.
Solar-GECO is designed to predict the power conversion efficiency (PCE) of perovskite solar cells by considering both the geometric structure of the perovskite absorber and the chemical composition of the transport layers and other device components. The model combines a geometric graph neural network (GNN) that encodes the atomic structure of the perovskite with language model embeddings that process the textual descriptions of the chemical compounds. This integrated approach allows Solar-GECO to capture both intra-layer dependencies and inter-layer interactions, providing a more comprehensive understanding of the device’s performance.
One of the key advantages of Solar-GECO is its ability to predict not only the PCE but also the associated uncertainty. This feature is crucial for guiding experimental efforts and optimizing the design of perovskite solar cells. The researchers demonstrated that Solar-GECO outperforms existing models, reducing the mean absolute error (MAE) for PCE prediction from 3.066 to 2.936 compared to the previous state-of-the-art model, semantic GNN.
The practical applications of this research are significant for the energy sector. By accurately predicting the performance of perovskite solar cells, Solar-GECO can help researchers and manufacturers quickly identify the most promising materials and architectures, speeding up the development and deployment of more efficient and cost-effective solar energy solutions. This, in turn, can contribute to the broader adoption of renewable energy and the transition to a more sustainable energy future.
Source: Li, L., Puel, J.-B., Carton, F., Barrit, D., & Giraldo, J. H. (2023). Solar-GECO: Perovskite Solar Cell Property Prediction with Geometric-Aware Co-Attention. Nature Communications.
This article is based on research available at arXiv.