Numerical investigation of thermal energy storage in wavy enclosures with nanoencapsulated phase change materials using deep learning

Abstract

The efficient storage and utilization of thermal energy remain critical challenges in advancing sustainable energy solutions, particularly in applications involving phase change materials. Nanoencapsulated phase change materials offer significant advantages, including compact dimensions, high specific surface area, superior thermal stability, and enhanced heat transfer performance, making them ideal candidates for thermal energy storage. However, accurately modeling the thermal behavior of these materials within complex enclosures, such as wavy structures, remains a computationally intensive and time-consuming challenge. To address this limitation, this study leverages deep learning techniques to precisely predict the thermal energy storage properties of nanoencapsulated phase change materials in wavy enclosures. Three different artificial neural network models were developed to simulate the thermal properties of the system, with each model incorporating varying input parameters and employing the Levenberg-Marquardt training algorithm. The outputs generated by the multilayer perceptron network models were compared against experimental data, demonstrating an excellent fit. Performance evaluations indicated that the developed models achieved exceptionally high prediction accuracy, with an average deviation of less than −0.65 %. The findings of this study highlight the potential of deep learning as a powerful predictive tool in thermal energy storage applications. By significantly reducing computational costs while maintaining high accuracy, this approach offers a transformative solution for optimizing energy storage system design.