Hierarchical variance analysis of solar cell production using machine learning and numerical simulations

Solar cells are an important source of renewable energy and have become a major industry with annual production capacities above 200 GW. The large amount of data produced during the solar cell production process calls for big data solutions and machine learning based approaches to improve quality and increase efficiency. In this work, we analyzed solar simulator data from 60,000 PERC solar cells using a hierarchical model based on machine learning, and compared the results to a theoretical model to extract missing information and determine the most likely input parameters for each produced solar cell. This approach enabled us to gain insight into the mechanisms that lead to the variance of conversion efficiency.