DEVELOPMENT OF AN AI-INTEGRATED PREDICTIVE MODELING FRAMEWORK FOR PERFORMANCE OPTIMIZATION OF PEROVSKITE AND TANDEM SOLAR PHOTOVOLTAIC SYSTEMS

Abstract

This study explores the integration of artificial intelligence into predictive modeling frameworks aimed at optimizing the performance of perovskite and tandem solar photovoltaic systems, technologies that have emerged as transformative solutions for enhancing solar energy efficiency and accelerating global decarbonization efforts. Perovskite solar cells have gained significant attention due to their tunable bandgaps, high defect tolerance, and cost-effective fabrication methods, while tandem architectures—particularly perovskite–silicon combinations—offer the potential to surpass the Shockley–quizzer efficiency limit. Despite these advantages, persistent challenges related to instability, ion migration, current matching, and spectral sensitivity hinder widespread adoption. Artificial intelligence has been increasingly applied across these domains, enabling high-throughput materials discovery, surrogate modeling for device physics, reinforcement learning for maximum power point tracking, and computer vision approaches for defect detection. This study followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines to ensure methodological rigor and transparency, systematically analyzing 240 peer-reviewed articles published between 2010 and 2025, which collectively accounted for more than 25,000 citations in the scientific literature. The evidence synthesized from these studies demonstrates that physics-informed machine learning provides a powerful means to balance mechanistic accuracy with data-driven adaptability, particularly for addressing nonlinearities in device behavior and forecasting challenges unique to tandem systems. Furthermore, international collaborative initiatives, data standardization efforts, and benchmarking protocols emerged as essential enablers for scaling AI applications from laboratory devices to field-deployed modules. The findings highlight that the convergence of perovskite and tandem device research with AI-driven predictive modeling establishes a robust pathway for improving energy yield, enhancing stability, reducing operational costs, and ensuring reliable integration into grid systems. This synthesis not only consolidates existing knowledge but also provides a structured foundation for advancing AI-enabled optimization strategies that are vital for the next generation of sustainable photovoltaic technologies.