Machine learning aided optoelectric characterization modelling and prediction of the IV parameters of perovskite solar cells with > 90% accuracy.

Luminescence imaging techniques assume a critical role in the evaluation of the durability and lifespan of solar cells. This study presents a novel empirical approach that capitalizes on simple machine learning-based linear regression (MLLR) to prognosticate the performance of perovskite solar cells (PSCs). Three types of MLLR models were developed: (a) Full width at Half maximum (FWHM)-based MLLR Model, (b) Colour Correlated Temperature (CCT)-based MLLR Model, and (c) Hybrid or FWHM-CCT-based MLLR Model. The proposed FWHM-based MLLR model achieves the best overall performance in fitting all current-voltage (IV) parameters. The CCT-based model, while applicable to all parameters, exhibits slightly lower accuracy compared to the FWHM-based approach; however, considering standalone analysis, CCT-based MLLR model is the best fit for estimation of Voc and FF parameters. This prognostication hinges on quantitatively assessing the FWHM using electroluminescence (EL) spectroscopy data. Notably, this investigation stands out by showcasing the potential of simple yet supervised machine learning in appraising IV characteristics of PSC through EL spectroscopy. Remarkably, the study attains an average predictive accuracy of beyond 90% for IV characteristics via EL spectroscopy facilitated by a supervised machine-learning model using both continuous and discontinuous datasets. The findings of this research underscore the potential of supervised machine learning as an innovative technique for approximating IV curve parameters of PSC, utilizing EL spectroscopy. This advancement holds significant ramifications for fortifying the assessment of PSCs in terms of reliability analysis and manufacturing efficiency.