Machine learning-integrated SCAPS-1D simulation for optimizing N719/BEHP-co-MEH-PPV bilayer dye-sensitized solar cells

Abstract Recent advances in dye-sensitized solar cells (DSSCs) have driven interest in novel absorber and transport layer combinations to overcome limitations of conventional designs, such as narrow absorption spectra and inefficient charge separation. This study presents a bilayer DSSC integrating N719 and BEHP-co-MEH-PPV dyes as a dual absorber system—implemented together for the first time. Sixty-four device configurations were constructed by varying hole and electron transport layers and back contact materials. SCAPS-1D simulations were employed to analyse device parameters, including doping levels, defect densities, layer thicknesses, resistances, and temperature effects. The optimal configuration—FTO/ZnO/N719/BEHP-co-MEH PPV/PEDOT:PSS/Pd—yielded a power conversion efficiency (PCE) of 7.52%, with a Voc of 1.31 V, Jsc of 8.29 mA.cm −2 , and a fill factor of 69.47%. ZnO and PEDOT:PSS outperformed other ETL/HTL combinations, suggesting improved charge transport and band alignment. Additionally, a machine learning-based regression model using a Decision Tree Regressor accurately predicted PCE with R 2 = 0.919 and MAPE = 11.17%, validating simulation insights and enabling rapid device optimization.