Numerical Investigation Using Machine Learning Process Combination of Bio PCM and Solar Salt for Thermal Energy Storage Applications

Abstract

TGA kinetic analysis can assess the thermal stability and degradation properties of PCMs by calculating activation energies and onset degradation temperatures, which are critical elements when developing optimal PCM composition and assessing long-term performance in thermal energy storage applications. In this study, we utilize a thermogravimetric analyzer to examine the thermal stability of both solar salt phase change material (i.e., commonly used in medium-temperature applications) (NaNO3 + KNO3) and a composite eutectic PCM mixture (i.e., PCM with 20% biochar). The activation energies of both the pure solar salt and composite solar salt PCM sample were evaluated using a variety of different kinetic models such as Kissinger–Akahira–Sunose (KAS), Flynn–Wall–Ozawa (FWO), and Starink. For pure PCM, the mean activation energies calculated using the KAS, FWO, and Starink methods are 581.73 kJ/mol, 570.47 kJ/mol, and 581.31 kJ/mol, respectively. Conversely, for the composite solar salt PCM sample, the calculated experimental average activation energies are 51.67 kJ/mol, 62.124 kJ/mol, and 51.383 kJ/mol. Additionally, various machine learning models, such as linear regression, decision tree regression, gradient boosting regression, random forest regression, polynomial regression, Gaussian process regression, and KNN regression models, are developed to predict the degradation behaviour of pure and composite solar salts under different loading rates. In the machine learning models, the mass loss of the samples is the output variable and the input features are PCM type, heating rate, and temperature. The machine learning models had a great prediction performance based on experimental TGA data, with KNN regression outperforming the other models by achieving the lowest RMSE of 0.0318 and the highest R2 score of 0.977.