Adaptive machine learning based control of multi-temperature, multi-module latent thermal storage ensembles

Abstract

Phase change material thermal energy storage systems can be used to alleviate disparities in renewable energy supply and demand. Performance of such systems can be enhanced using a cascade of multiple phase change materials but efficiency benefits currently rely on precise optimization of design parameters. Therefore, there is motivation to develop storage technologies which provide greater flexibility for applications with dynamic operating conditions. In this study, a novel system is proposed called a multi-temperature, multi-module ensemble which consists of multiple phase change materials combined in series and parallel configurations. The system is optimized and controlled by an artificial neural network trained using a multi-objective optimization scheme. The performance for the novel system is compared to two baseline systems: a single phase change material system and a cascaded system. Across a wide range of fixed operating conditions, the results show that the ensemble reduces the rate of exergy destruction by 60% compared to the single module system and achieves a 10 °C improvement in performance for matching target outlet temperature compared to the cascaded system. Over a dynamically changing load profile, the results show that the multi-temperature, multi-module ensemble with artificial neural network controller is the preferable design as it achieves the excellent exergy efficiency of a cascaded system while maintaining the flexibility needed to respond to dynamically changing operating conditions. Furthermore, this novel approach widens the potential applications of latent thermal energy storage by providing a means to standardize manufacturing to drive down system cost and allowing the system to adapt to unforeseen design parameters.