How thermal storage modeling affects integrated energy systems simulation: Lumped vs. stratified models under fast and high-capacity thermal loads

Accurate modeling of thermal energy storage (TES) is crucial for predicting energy flows in integrated energy systems, especially in applications like Domestic Hot Water (DHW) production, where high-resolution simulations are needed to capture temperature-dependent energy performances and control interactions under fast and high-capacity thermal loads. However, many studies employ zero-dimensional (0-D) lumped models due to their simplicity and reduced computational time, often overlooking the significant impact of storage modeling on energy exchange evaluation. In this context, the paper offers a comparison between the energy results and computational times of the 0-D lumped model approach and the one-dimensional (1-D) stratified finite volume model approach for TES, encouraging a sensitivity analysis to determine the optimal number of TES nodes for achieving a balance between computational efficiency and modeling accuracy. The TES model has been validated with experimental data to assess its capability in simulating temperature dynamics and the accuracy of heat transfer correlations for submerged coils. The experimental campaign performed on a typical stratified water tank used in residential applications (i.e., 2-m height) shows that the 1-D model with 10 nodes significantly reduces the deviations observed with the 0-D model, albeit with an approximate threefold increase in computational time. Further increasing the number of nodes does not appreciably reduce the deviation but only increases computational time. Both models were then used to estimate the seasonal performance of a solar-assisted heat pump system for DHW production in a typical Mediterranean single-family house, including TES, a heat pump, a solar collector, and a backup generator. The 0-D and 1-D models showed deviations in energy employed by the thermal solar collector, heat pump, and backup generator up to 29%, 70%, and 53%, respectively. Additionally, the 0-D model failed to accurately replicate the dynamics of controls and component activation, as it could not properly evaluate the thermal inertia experienced by temperature sensors used for generator control. For our case study, we conclude that using 5 to 10 nodes offers, the 1-D model provides a reasonable trade-off for high-resolution seasonal dynamic simulations with control-oriented purposes. Provided that each specific case requires its own sensitivity analysis to choose the best number of nodes, we suggest performing a convergence analysis on the main energy quantities of interest and computational time. Additionally, in similar energy systems under fast and high-capacity thermal loads, we suggest performing a sensitivity analysis on the simulation timestep to properly capture the dynamics of thermal load and controls.