Accelerating the Design of Double-Absorber Solar Cells: From Surrogate Model-Assisted Reinforcement Learning and...

Area of Science:

• Materials Science
• Renewable Energy
• Computational Modeling

Background:

• Lead-free double-absorber perovskite solar cells offer environmental advantages and broad light absorption.
• Their complex multilayer structures pose significant computational challenges for traditional optimization methods.

Purpose of the Study:

• To develop an automated framework for optimizing perovskite solar cell design.
• To establish an efficient and generalizable paradigm for photovoltaic device discovery and validation.

Main Methods:

• Integration of SCAPS-1D simulation, multilayer perceptron (MLP) surrogate modeling, metaheuristic algorithms, and reinforcement learning (RL).
• Utilized Latin hypercube sampling for MLP training and proximal policy optimization (PPO) for RL-based efficiency enhancement.
• Employed simulated annealing, particle swarm optimization, and Grey Wolf Optimizer (GWO) for direct dimensional optimization.

Main Results:

• The MLP model achieved high accuracy (R² > 0.95) in predicting device performance.
• The PPO RL agent reached 27.41% power conversion efficiency (PCE) rapidly.
• Metaheuristic algorithms efficiently approached the target PCE, with GWO showing superior average PCE.
• Transfer learning successfully adapted the model to new perovskite structures, enhancing prediction accuracy for PCE, J_SC, and FF.

Conclusions:

• The developed automated framework enables systematic optimization of perovskite solar cells.
• This approach offers a generalizable and efficient paradigm for intelligent photovoltaic device design and material discovery.