A dual-track, physics-informed framework for predicting TENG open circuit voltage and short circuit current.

Area of Science:

• Materials Science
• Electrical Engineering
• Energy Harvesting

Background:

• Triboelectric nanogenerators (TENGs) show promise for self-powered sensing and energy harvesting.
• Predicting TENG output is difficult due to device heterogeneity and varied reporting standards.
• Standardized prediction methods are needed for reliable TENG design and comparison.

Purpose of the Study:

• To develop a robust framework for predicting TENG performance (V_oc, I_sc).
• To establish actionable design rules for optimizing TENGs.
• To enable reproducible comparisons across diverse TENG studies.

Main Methods:

• A dual-track approach: a closed-form surrogate model based on Thevenin load matching and a machine learning predictor.
• Training the machine learning model using leave-one-study-out validation on a harmonized corpus of 20 studies (443 conditions).
• Incorporating uncertainty quantification and study-level robustness checks.

Main Results:

• Accurate prediction of open-circuit voltage (V_oc) and short-circuit current (I_sc) with calibrated prediction intervals.
• Estimation of key internal parameters like resistance, capacitance, and charge density.
• Identification of design rules, such as optimal load matching to internal resistance.

Conclusions:

• The framework enables physics-guided prediction and design of TENG systems.
• It facilitates reproducible performance comparisons across different TENG studies.
• The approach supports the advancement of TENGs for practical applications in energy harvesting and sensing.