To stabilize power grids with high renewable energy penetration, steam accumulators (SA) offer flexible operational support for power systems; however, their transient behavior is not adequately captured by conventional models. This study proposes a non-equilibrium dynamic model that incorporates direct-contact condensation mechanisms and evaluates phase transition kinetics based on phase thermodynamic properties, thereby improving the representation of actual physical processes and eliminating the reliance on empirical relaxation parameters. The model is validated using three independent experimental datasets, demonstrating significantly improved accuracy compared to a classical model, with average reductions of 17.14% in RMSE and 16.91% in MAPE for pressure predictions. The parametric analysis reveals the influence mechanisms of initial parameters and charging/discharging conditions on SA dynamic characteristics. During charging, increasing the steam injection flow rate accelerates pressurization by 62.7% but intensifies the self-balancing pressure drop to 20.9%, while higher temperatures reduce the final liquid mass increment by 10.9%. Elevated initial water content prolongs the pressurization duration and increases relative steam loss to 38.9% during self-balancing stage, whereas a higher initial pressure enhances the pressure rise rate but reduces the final water content. During discharging, greater initial water content mitigates the pressure drop rate, while elevated initial pressure accelerates pressure decline. Steam discharge achieves 73.3% higher output power compared to water discharge, accompanied by a 6.35-fold faster pressure decline rate. This work elucidates coupled heat-mass transfer mechanisms in SAs and enables quantitative optimization of capacity sizing, operating set-points, and discharge media selection for balancing energy density, efficiency, and output power. • Non-equilibrium model with direct-contact condensation and phase transition kinetics • Physically realistic modeling improves pressure prediction accuracy by 17.14% in RMSE • Charging flow rate and initial water content enhanced self-balancing pressure drop • Steam discharge has 6.35-fold faster pressure drop but 73.3% higher power than water
Zhang et al. (Fri,) studied this question.