Abstract
Thermal Energy Storage (TES) platforms can help balance the difference between energy consumption and supply. Phase change materials (PCMs) can enhance the performance of TES platforms by improving their resilience and reliability, as the high latent heat values of PCMs allow for more compact form factors. Inorganic PCMs have higher latent heat values than organic PCMs, but their reliability is often compromised due to their requirement for a high degree of supercooling for initiating nucleation for freezing. The “Cold Finger Technique (CFT)” can mitigate this issue by leaving a small portion of the PCM un-melted during the melt cycle, facilitating spontaneous nucleation during the freezing cycle. This study uses machine learning (ML) techniques to leverage the effectiveness of CFT by training an Artificial Neural Network (ANN) model to predict the time required to reach a designated melt-fraction of the PCM with outstanding accuracy. However, the fidelity of the dataset used to train the ANN algorithm can impact the accuracy of the predictions. The prediction errors were in the order ≤ 10% of the time required to reach the desired melt-fraction and are drastically bigger at 99% and 100% melt-fractions.