Hanbat Researchers Use AI to Optimize Thermal Energy Storage

Researchers at Hanbat National University have developed a physics informed neural network framework to speed design optimization for latent heat thermal energy storage systems, the university said in a press release. The study was made available online May 5, 2026, and published July 30 in Volume 167 of the Journal of Energy Storage.

The team built a lab setup, validated a computational fluid dynamics model against measurements, then used 15 high fidelity simulations to train the neural network. The model combines embedded physical equations with learned heat transfer coefficients and a response surface model for geometric effects.

The researchers paired the digital twin with the NSGA-II optimization algorithm to search designs that maximize discharged heat and average power while minimizing pumping power. In numerical tests, the AI model reproduced CFD predictions and found a design that matched the strongest baseline thermal performance while reducing pumping power.

The paper reports that flatter pipes performed better within the tested design range. The researchers said the approach could also be applied to thermal management in electric vehicle batteries, data centers, cold chain logistics, and solar thermal systems.