State of charge estimation using regression models in a novel photovoltaic thermal storage system with macro-encapsulated phase change material
Abstract
Photovoltaic heat pump systems (PV-WP Systems) are becoming a standard for low-carbon building heating. In this framework, this work aims to optimize photovoltaic (PV) systems coupled with heat pumps (WP) by integrating a novel high-density thermal storage unit to address the mismatch between solar energy production and heating demands. Through a pilot setup featuring an 800 L storage tank equipped with multiple sensors, comprehensive data was collected over the 2022/2023 heating seasons in Switzerland. This dataset formed the base for the development and validation of state of charge (SoC) estimation models for latent thermal energy storage (LTES) systems. Challenges in defining SoC due to temperature differentials within the storage were addressed using both energy balance and machine learning approaches. The machine learning approach, leveraging regression techniques and ensemble algorithms, proved particularly effective, achieving a prediction accuracy with a deviation of less than 2.06 kWh for 95% of data points with a total storage capacity of 45 kWh. This approach enabled adaptive SoC predictions, enhancing the operational efficiency of the storage system without additional hardware. Further work will focus on redefining of these models to improve accuracy, explore scalability across different system configurations, and reduce computational demands to facilitate integration into existing energy management systems. This research indicates significant potential for advancing thermal energy storage technology in PV-WP systems, contributing to more sustainable building heating solutions.