No abstract

Energy storage systems are key elements for enabling the design of MicroGrids in buildings, specially to deal with stochastic renewable energy resources and to promote peak shifting. However, inaccuracies in the batteries' mathematical models due to temperature and ageing effects can reduce the performance of a MicroGrid system. To tackle these uncertainties, this article presents a two-level hierarchical model predictive controller empowered with a data-driven algorithm for real-time model identification of Lithium-ion batteries. The objective is to enhance their state of charge estimation and to make their maximum use without damaging them. The results demonstrate that it improves up to three times the accuracy of state-of-charge estimation and increases about 3% the annual building MicroGrid selfconsumption rate. Furthermore, the division of the building MicroGrid energy management system into two hierarchical levels soften the drawbacks arise from the inaccuracies of day-ahead data prediction while reducing the computational cost. The proposed architecture guarantees higher energetic autonomy indexes than a conventional rule-based controller in all scenarios under study.

No abstract