Researchers from the University of Alberta, including Meraj Hassanzadeh, Ehsan Ghaderi, Fatemeh Fatahi, and Mohamad Ali Bijarchi, have developed a novel approach to model thermal energy storage systems using phase change materials (PCMs). Their work, published in the Journal of Computational Physics, focuses on improving the efficiency of these systems, which are crucial for sustainable energy management and grid stability.
Thermal energy storage (TES) systems using PCMs absorb and release heat during the phase transition from solid to liquid and vice versa. These systems are integrated with finned heat exchangers and operate under transient forced convection with cooling air. However, modeling these systems is challenging due to the complex dynamics of the moving phase boundary and the significant computational resources required for conventional numerical methods.
The researchers developed a Physics-Driven Deep Learning (PDDL) framework to overcome these challenges. This framework uses three specialized deep neural networks operating in parallel to predict the solid phase temperature field, fin temperature distribution, and the moving phase boundary position. The networks are constrained by the governing physical laws of energy conservation and interface conditions.
The PDDL framework was validated against established analytical benchmarks and demonstrated exceptional accuracy in predicting interface evolution and temperature distributions across various aspect ratios. It successfully captured the influence of geometrical parameters on solidification rates and thermal performance without requiring mesh regeneration, a significant advantage over conventional methods.
The practical applications of this research for the energy sector are substantial. The PDDL framework provides an efficient computational paradigm for optimizing PCM-based TES systems. It can help design more effective thermal energy storage solutions, which are essential for integrating renewable energy sources into the grid and improving overall energy efficiency. Additionally, the framework is extensible to three-dimensional configurations and multi-material composites, presenting further potential for advanced thermal energy system design.
This research represents a significant step forward in the modeling and optimization of thermal energy storage systems. By leveraging the power of deep learning and physics-driven constraints, the researchers have developed a tool that can enhance the performance and efficiency of these critical energy technologies.
This article is based on research available at arXiv.