Multi-scale hybrid composites of erythritol/diatomite/GNP for enhanced thermal energy storage: experimental and machine learning optimization.

Thermal energy storage (TES) using phase change materials (PCMs) is a critical technology for balancing energy supply and demand. Among sugar alcohols, erythritol stands out due to its high latent heat; however, its practical application is hindered by its low thermal conductivity and leakage during phase transition. In this study, a multi-scale hybrid approach is developed to overcome these limitations by integrating erythritol with a micro-porous diatomite framework and graphene nanoplatelets (GNPs) as thermal enhancers. The methodology combines experimental fabrication through the impregnation method with advanced machine learning (ML) optimization to identify the ideal composition. Results demonstrate that the composite with 4 wt% GNP and 40 wt% diatomite achieved a significant thermal conductivity enhancement of 261% (reaching 2.64 W/m·K) while maintaining excellent shape stability. ML models, specifically )MLPNN, successfully predicted thermal performance with high accuracy (R² > 0.98). These findings suggest that the optimized hybrid composite provides a high-performance and cost-effective solution for mid-temperature thermal energy storage systems.