Comparative analysis of aggregate battery models to characterize the flexibility of electric water heaters

Due to the fast deployment of renewable energy sources in the power system, an increasing amount of flexibility is required. Residential thermostatically controlled loads, such as electric water heaters, are good candidates to provide this flexibility at the demand side. However, the provision of flexibility with such appliances involves a large-scale control problem which poses computational challenges. To address these issues, a recent line of research has been dedicated to the development of aggregate models, which allow to represent a large group of thermal appliances as one entity. Within these aggregate representations, virtual battery models are particularly useful, as they provide a clear view of the available flexibility and can easily be embedded in model predictive control frameworks. In this paper, relevant quality metrics for virtual battery models
are first identified. Based on a realistic group of electric water heaters, these metrics are then used to compare a series of virtual battery models developed in recent literature. The strength and weaknesses of the studied models are identified. The strong trade-off between the degree of conservativeness of the models and the reliability of their associated control actions is highlighted and quantified. It is shown that the virtual battery model and the appliances coordination strategy
should be designed jointly to produce a reliable control approach. The findings of this work can be used as a starting point for the design of new virtual battery modelling approaches.