Physics-Guided Surrogate Modeling for Cross-Architecture Normalized Power-Density Optimization in Perovskite Solar Cells.

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος:	Physics-Guided Surrogate Modeling for Cross-Architecture Normalized Power-Density Optimization in Perovskite Solar Cells.
Συγγραφείς:	Mourched, Bachar¹ (AUTHOR) bachar.mourched@aum.edu.kw, Abdallah, Mariam² (AUTHOR), Vrtagić, Sabahudin¹ (AUTHOR)
Πηγή:	Journal Européen des Systèmes Automatisés. Jun2026, Vol. 59 Issue 6, p1729-1745. 17p.
Θεματικοί όροι:	Photovoltaic power generation, Solar cells, Deep learning, Solar cell design, Prediction models, Mathematical optimization, Computer simulation
Περίληψη:	A physics-guided deep learning surrogate framework is developed for rapid prediction and optimization of Dmax in perovskite solar cells (PSCs). Here, Dmax is defined as a simulated maximum normalized power-density indicator, not as experimentally measured power conversion efficiency. COMSOL drift-diffusion simulations generated 2,000 device configurations across two architectures: TiO₂/MAPbI₃/Spiro-OMeTAD and ZnO/CsFAPbI₃/PTAA. A deep neural-network surrogate was trained using physics-guided geometric and material descriptors, achieving a test R² of approximately 0.97 while reducing evaluation time from hours per simulation to milliseconds per prediction. Sensitivity and SHAP analyses identified absorber thickness as the dominant factor controlling Dmax, with additional architecture-dependent effects from transport-layer scaling and material properties. Grid-proximal virtual screening identified surrogate-predicted candidate design regions. Independent COMSOL validation showed strong agreement near the TiO₂/MAPbI₃/Spiro-OMeTAD candidate region and larger local deviations near the ZnO/CsFAPbI₃/PTAA predicted optimum. The framework provides a practical route for simulation-driven normalized power-density optimization and accelerated photovoltaic device design. [ABSTRACT FROM AUTHOR]
Βάση Δεδομένων:	Supplemental Index

καταχωρήστε σχόλιο πρώτοι!