Design optimization of large-scale bifacial photovoltaic module frame using deep learning surrogate model

Area of Science:

• Materials Science
• Mechanical Engineering
• Renewable Energy Systems

Background:

• Increasing wafer sizes in solar cells lead to larger, heavier photovoltaic (PV) modules.
• Module weight can cause deflection, potentially reducing PV cell efficiency.

Purpose of the Study:

• To develop an optimal frame design for large-scale bifacial PV modules to mitigate deflection.
• To identify frame design factors that minimize both deflection and weight.

Main Methods:

• Constructed a finite element (FE) model for large-scale bifacial PV modules.
• Trained a deep neural network (DNN)-based FE surrogate model using FEA datasets.
• Improved the model with Bayesian optimization and k-fold validation.
• Generated 1 million datasets to predict optimal frame design factors.

Main Results:

• The trained FE surrogate model demonstrated high accuracy with low error rates (MAPE < 0.0020, R² > 0.9962).
• Identified optimal frame design factors (a=1.5, b=13.7, c=1.5, d=3.0, e=4.3) to minimize deflection and weight.
• Achieved a 9.6% reduction in deflection and a 12.8% increase in weight compared to existing designs.

Conclusions:

• The FE surrogate model accurately predicts FEA results for PV module deflection and weight.
• The study provides a basis for designing lighter, more deflection-resistant PV module frames.
• Optimized frame designs can enhance the efficiency and longevity of large-scale bifacial PV modules.