Clustering-driven design and predictive control of hybrid PV-battery storage systems for demand response in energy communities

The integration of renewable energy sources and participation in demand response requires advanced modelling and control strategies to enhance building-grid interaction. This study presents a comprehensive methodology for selecting typical days and evaluating how controllable building thermal loads influence the design and operation of grid-supportive technologies, specifically photovoltaic (PV) and battery storage systems. Typical days are identified through dynamic time warping (DTW) and hierarchical clustering approaches, supported by six internal validation metrics. Grey-box and regression models are employed to predict building energy consumption, while PV and battery models assess system performance. A two-level Model Predictive Control (MPC) framework is employed to optimize the buildings demand and coordinate the operation of grid-supportive technologies. At the first level, a distributed MPC algorithm manages thermal loads in individual buildings to enable demand response. At the second level, a supervisory MPC optimizes the operation of the hybrid PV-battery storage system to achieve targeted grid flexibility. The case study considers a virtual community in Varennes, Québec, consisting of institutional and residential buildings. Through efficient thermal load management, the methodology shows that community peak demand can be reduced by over 40% compared to current operational practices, and the required capacity of grid-supportive systems can be reduced by up to 26% in a worst-case scenario analysis.