Predictive modeling of thermal conductivity in PCM composites using artificial Intelligence

Phase change materials (PCMs) are widely used in thermal energy storage (TES) systems due to their high latent heat capacity, yet their inherently low thermal conductivity restricts practical performance. While conductive fillers have been employed to improve heat transfer, identifying suitable filler–PCM systems remains a challenging and resource-intensive process. In this study, we apply and compare several machine learning models, including Linear Regression, Support Vector Regression (SVR), Random Forest, Gradient Boosting, and Artificial Neural Network (ANN), to predict the thermal conductivity of PCM composites. The models were trained on an extensive dataset compiled from the literature and assessed using statistical metrics and residual analyses. Among them, the ANN model achieved the highest predictive accuracy (R² ≈ 0.97) with residuals closely centered around zero, demonstrating strong generalization capability. The analysis further reveals that filler concentration and testing temperature are the most influential variables governing thermal conductivity. These findings highlight the potential of machine learning approaches as efficient computational tools for guiding the design and optimization of PCM composites, ultimately supporting their integration into advanced thermal management and sustainable energy applications.