Hanbat National University Researchers Reveal Physics-informed AI for Rapid Optimization of Thermal Energy Storage Systems

The proposed physics-informed neural network framework enables rapid, autonomous design optimization of latent heat thermal energy storage systems

CHUNGCHEONG PROVINCE, South Korea, Aug. 17, 2026 /PRNewswire/ -- Decarbonizing the global building sector requires efficient thermal energy storage to manage substantial heating and cooling demands. Latent heat thermal energy storage (LHTES) systems using phase change materials are emerging as a key solution because of their high energy storage density and ability to release thermal energy at a near-constant temperature, with previous studies suggesting potential energy savings of up to 45%.

However, optimizing LHTES systems is challenging because physical experiments are largely constrained to laboratory scales, while computational fluid dynamics (CFD) simulations—despite their ability to model complex coupled heat-transfer and fluid-flow processes across scales—are too computationally expensive and time-consuming for large-scale design optimization.

In a new study, a collaborative team of researchers from the Republic of Korea, led by Assistant Professor Joo Hyun Moon from the Department of Building Systems Engineering at Hanbat National University in South Korea, has developed a hybrid physics-informed neural network (PINN) framework for optimization of LHTES systems. Their study was made available online on May 05, 2026, and published in Volume 167 of the Journal of Energy Storage on July 30, 2026.

"Our framework integrates the governing laws of physics into an AI model, overcoming the limitations of slow and computationally expensive conventional simulations. It models the heat-release and solidification behavior of wax-type phase change materials, which act like thermal batteries by absorbing and releasing large amounts of heat. This enables us to rapidly explore and evaluate tens of thousands of potential system designs," explains Dr. Moon.

To develop the data-driven PINN, the researchers first created a high-fidelity ground-truth dataset that captures the physics of the LHTES system. For this, they developed a laboratory-scale LHTES setup and validated its CFD model against experimental measurements.

The validated CFD model generated a sparse dataset of 15 high-fidelity simulations to train a PINN that embeds a zero-dimensional physical model and learns case-specific heat-transfer coefficients. A response surface model then captures geometric effects, forming a fast digital twin capable of predicting unseen designs and operating conditions within the studied design range.

Coupled with a Non-dominated Sorting Genetic Algorithm II (NSGA-II), the digital twin optimizes designs by maximizing discharged heat and average power while minimizing pumping power.

In numerical experiments, the PINN accurately reproduced CFD predictions and enabled rapid, autonomous design optimization. The optimized design matched the best baseline's thermal performance while significantly reducing pumping power, with flatter pipes proving more favorable. Beyond buildings, this physics-informed approach could support thermal management in electric-vehicle batteries, data centers, cold-chain logistics, and solar thermal systems. Earlier studies suggest that smarter latent-heat-storage control could also reduce electricity costs by more than 70%.

"Our approach enables autonomous exploration of the continuous LHTES design space within the studied range, helping engineers develop more efficient systems while reducing energy use and carbon emissions," concludes Dr. Moon.

Reference
Title of original paper: Physics-informed neural networks for multi-objective design optimization of latent heat thermal energy storage systems
Journal: Journal of Energy Storage
DOI: https://doi.org/10.1016/j.est.2026.122514 

About the institute
Website: https://www.hanbat.ac.kr/eng/

Media Contact:
Seyoung Jang
82-42-828-8461
[email protected] 

SOURCE Hanbat National University