Abstract
Organic-inorganic metal halide perovskite solar cells are renowned for their extensive solution processability, although the production of uniformly crystalline perovskite films can necessitate intricate deposition methods. In our study, we harmonized Shockley diode-based numerical analysis with machine learning techniques to extract the device characteristics of perovskite solar cells and optimize their photovoltaic performance in light of the experimental variables. The application of the Shockley diode equation facilitated the extraction of photovoltaic parameters and the prediction of power conversion efficiencies, thus aiding the understanding of device physics and charge recombination. Through machine learning, specifically Gaussian process regression, we trained models on current-voltage curves sensitive to variations in fabrication conditions, thereby pinpointing the optimal settings for enhanced device performance. Our multifaceted approach not only clarifies the interplay between experimental conditions and device performance but also streamlines the optimization process, diminishing the need for exhaustive trial-and-error experiments. This methodology holds substantial promise for advancing the development and fine-tuning of next-generation perovskite solar cells.