Data-driven quantification and aggregation of demand-side flexibility for symmetrical bidding in energy balancing markets

Modern power grids are transitioning towards a net-zero framework with carbon-neutral, and weather-dependent generation, necessitating substantial demand-side flexibility to manage the variability in generation and consumption. This paper introduces a data-driven approach to quantify and aggregate demand-side flexibility for effective participation in the energy balancing market. The methodology applies machine learning algorithms to characterize and quantify flexibility from various household appliances, including shiftable loads, thermostatic devices, and battery storage systems, while considering dynamic usage behavior. The quantified flexibility is aggregated per appliance type across households, generating flexibility profiles for both low and high comfort disturbance zones, enabling a detailed assessment of flexibility potential within a residential community. The aggregated flexibility is then optimized using a sequential Mixed-Integer Linear Programming (MILP) model to maximize utilization while adhering to operational and market constraints, including minimizing user discomfort and fulfilling symmetrical bidding requirements for both up- and down-regulation over the next four-hour pool. Two case studies are presented: the first demonstrates the flexibility quantification of a single household, illustrating dynamic user behavior in appliance usage, while the second presents the aggregated flexibility of multiple households, showcasing community-level potential. The results validate the approach's effectiveness in quantifying and aggregating demand-side flexibility, achieving an average utilization of 82 % from the low comfort zone, thereby reducing user discomfort and facilitating strategic participation in balancing markets.