Data-driven approaches to sustainable phase change material selection in latent heat storage systems

This research investigates sustainable phase change materials (PCMs) for latent heat thermal energy storage systems using data-driven machine learning models. Activated biochar is incorporated as a support material to improve the PCM’s thermal conductivity and leak resistance during phase transition, with multi-walled carbon nanotubes further enhancing thermal conductivity. The phase change kinetics of the PCM composites are analyzed using Kissinger, Ozawa, and Starink kinetic models, revealing that both activated biochar and carbon nanotubes significantly reduce activation energy. Several machine learning models including Linear Regression, Polynomial Regression, Decision Tree, and Random Forest regression are developed to predict differential scanning calorimetry (DSC) data of the composites. Results indicate that Linear Regression performed inadequately, with an R² of 0.359 and RMSE of 0.153, failing to predict the DSC data complexity. Polynomial Regression improved prediction accuracy, achieving an R² of 0.875 and RMSE of 0.067. The Decision Tree model achieved high accuracy (R² of 0.9998 and RMSE of 0.003), effectively capturing non-linear patterns. Random Forest regression outperformed all models, with an R² of 0.9998 and RMSE of 0.0025, demonstrating the robustness and reliability of ensemble methods for predicting PCM composite behavior. As a result, a random forest regression model was successfully used to predict experimental DSC data, providing a better understanding of the thermophysical properties and phase change kinetics of PCM composites.