Resumen de Performance analysis and control optimization of a solar-driven seasonal sorption thermal energy storage system
Alicia Crespo Gutierrez
Heat is the world¿s largest energy end-use. In particular, the building sector is responsible of around 40 % for the consumed heat. Renewable sources are a good solution to mitigate climate change. Nevertheless, the intermittence of renewable sources calls for the need of energy storage systems. Solar thermal collectors coupled with seasonal thermal energy storage (STES) are a good solution to reduce the fossil fuels consumption in climates with high solar irradiation in summer and high space heating demand in winter. Sorption thermal energy storage (TES) is the more suitable technology for STES due to its nearly zero thermal losses during the storage period and its high energy density at material level. Nevertheless, the operation of a sorption STES integrated into a building heating system is not straightforward and must be studied in detail. A non-optimal operation of the system based on transient weather conditions and thermal building demands may lead to low system efficiency. This PhD thesis aims to analyse and enhance the performance of a solar-driven seasonal sorption TES (SDSSTES) system integrated into a building through different control strategies and system designs. The system was composed of solar collectors, a stratified water tank, a boiler, a sorption STES, and its low-temperature heat source (LTHS). Operating the system with an optimized rule based control (RBC) strategy allowed to minimize the operational costs using a lower volume of sorption TES. Moreover, the energy density of the sorption TES was highly impacted by the weather conditions, and by the type and availability of LTHS. The results proved the technical feasibility of the SDSSTES in Central and North Europe. In spite of the low temperatures in winter, the use of winter solar heat was enough to assist the discharge of the sorption TES. However, energy densities increased by 23 % assuming a constant heat source (e.g. geothermal energy). Better results in terms of operational costs were obtained by operating the system with deep reinforcement learning (DRL), in comparison to the optimized RBC strategy. Indeed, the use of DRL allowed operating the system during winter in a near-global optimum. Nevertheless, the development and implementation of a DRL algorithm require high programming skills and long computational training times.
