Revealing flexibility: A cluster-based, data-driven model for unveiling electricity consumption adaptability in households

The integration of residential demand-side flexibility into power system operations is essential for addressing grid stress during peak hours and enhancing the deployment potential of decentralized energy systems. This paper introduces a data-driven framework for identifying, evaluating, and aggregating household-level flexibility using only smart meter data. The method combines bottom-up clustering of daily electricity usage patterns with top-down, system-defined peak period detection based on dynamic thresholds. A second-stage clustering extracts feasible historical load-shifting scenarios conditioned on behavioral similarity. Each scenario is evaluated using a multi-criteria matrix that includes peak demand reduction, behavioral deviation, scenario occurrence probability, and electricity cost under dynamic pricing. The approach is demonstrated using two datasets and supports both exhaustive scenario enumeration and scalable aggregation strategies. Results show that the framework captures a spectrum of realistic flexibility options and reveals trade-offs between peak shaving potential and behavioral plausibility. By quantifying flexibility in terms of both technical effectiveness and behavioral deviation, this study offers a replicable and operationally relevant tool for grid operators and aggregators to design targeted, equitable, and dynamic demand-side management interventions.