Machine-Learning-Powered Breakthrough in Modular Thermal Energy Storage Design
A new study demonstrates how combining CFD simulations with machine-learning surrogate models can reduce thermal energy storage (TES) design time by more than 99%, marking a major step in enabling scalable renewable-energy systems.
Thermal energy storage (TES) is emerging as a key enabler of renewable-energy integration, particularly in concentrated solar power (CSP) and hybrid energy systems. A 2025 research study titled “Computationally Effective Machine Learning Approach for Modular Thermal Energy Storage Design” introduces a groundbreaking methodology that pairs classical computational fluid dynamics (CFD) with machine-learning (ML) models to accelerate TES design.
The researchers developed an ML surrogate model trained on CFD simulation data to predict the thermal discharge behavior of concrete-based TES modules under various inlet temperatures and flow conditions. The resulting model achieves high accuracy while reducing computational cost by over 99%, allowing real-time evaluation of complex TES layouts.
This dramatic increase in computational efficiency enables designers to evaluate modular TES systems containing multiple interconnected units — a task previously limited by the high cost of multi-module CFD simulations. The approach has promising implications for large-scale renewable-energy plants, district heating networks, and industrial waste-heat recovery.
As renewable-energy deployment accelerates globally, ML-enabled TES design could play a central role in enhancing the flexibility and resilience of modern energy infrastructures.
Key Takeaways
- ML surrogate models can reduce TES simulation time by over 99%.
- Modular TES systems become significantly easier and faster to design.
- The method enables scalable deployment of large-scale renewable-energy storage solutions.
References
Singh, D., Rugamba, T., Katara, H., & Grewal, K. S. (2025). Computationally effective machine learning approach for modular thermal energy storage design. Applied Energy, 377, 124430.
https://www.sciencedirect.com/science/article/pii/S0306261924018130
Rahjoo, M., Rojas, E., & Goracci, G. (2024). Data-Driven and Machine Learning-Enabled Design and Optimization of Solid-Based Thermal Energy Storage Units. Heliyon (preprint).
