Optimized Three-Layer Carbon Electrode Architecture for High-Performance CsPbI3 Perovskite Solar Cells

Perovskite solar cells (PSCs), particularly cesium lead iodide (CsPbI3)-based ones, offer low fabrication costs, high efficiency, and potential for scalable production. However, a significant drawback is their instability, especially under moisture and high-temperature conditions, which can lead to degradation and reduced long-term performance. To overcome these issues, this paper presents a novel approach for optimized three-layer carbon electrode (CE) architecture for high-performance CsPbI3 PSCs. The proposed method integrates the capabilities of leaf in wind optimization (LWO) algorithm and progressive graph convolutional networks (PGCNs), forming a model termed LWO-PGCN. The main goals of the proposed technique are to develop the power conversion efficiency (PCE) and stability of PSCs. The LWO algorithm performs global optimization of structural parameters, while the PGCN model predicts photovoltaic performance by capturing complex relationships between device parameters, enabling efficient identification of high-performance device configurations. The LWO algorithm is employed to optimize perovskite layer thickness, hole transport material (HTM) layer thickness. The PGCN is used to predict the PCE with high accuracy. The proposed method demonstrates better performance than existing techniques, including deep neural network (DNN), probabilistic neural networks (PNN), and artificial neural network (ANN), as evaluated and compared using the MATLAB platform. The proposed method achieves a higher PCE of 29.9%, short-circuit current density (J sc) of 25.84 mA/cm2, open-circuit voltage (V oc) of 1.32 V, and fill factor (FF) of 74.89%, indicating superior photovoltaic performance compared with existing methods. The proposed LWO-PGCN model significantly enhances the stability and efficiency of CsPbI3-based PSCs, outperforming existing methods in photovoltaic performance.