Machine learning-guided analysis of CIGS solar cell efficiency: Deep learning classification and feature importance evaluation

Abstract

The increasing sensitivity of thin-film solar cells to variations in design parameters is becoming more pronounced with ongoing advancements in material science and device engineering. Even minor deviations in these parameters can significantly impact the performance of the solar cell devices. Therefore, an in-depth investigation of such variations is crucial to advancing the efficiency and reliability of thin-film solar cell technology. This paper presents an innovative design methodology by combining numerical simulations with Machine Learning (ML) techniques to explore and analyze the critical design parameters of CIGS thin-film solar cells. Specifically, accurate numerical simulations, incorporating defects and appropriate charge transport mechanisms in different layers, are utilized to simulate the current–voltage (I–V) characteristics. The study thoroughly examines the effects of design parameter variations and the role of high-efficiency absorber and buffer layer materials on energy conversion performance. Several ML algorithms are employed to evaluate and classify the most influential design parameters impacting key figures of merit, such as power conversion efficiency (PCE), open-circuit voltage (V_OC), and short-circuit current density (J_SC). The results highlight the significant impact of the buffer and absorber layers properties on the overall efficiency. These parameters are also crucial for predicting and optimizing PCE. Furthermore, ML models are leveraged to identify the optimal design parameters for maximizing each Figure-of-merit (FoM).