A stochastic model for dynamic reconfiguration of multi-microgrid networks under demand and supply uncertainties

Efficiently operating a single microgrid (MG) is increasingly challenging due to volatile electricity demand and intermittent renewable generation. Traditional static networks often fail to adapt to these fluctuations, compromising reliability. Incorporating these uncertainties into planning is essential for developing resilient optimization models that can withstand the stochastic nature of decentralized energy systems. This study proposes a dynamic reconfiguration strategy for interconnected microgrids that reroutes households based on real-time supply and demand. A stochastic nonlinear optimization model was developed to maximize load factors and flatten peaks while accounting for current-dependent power and distribution losses. The Sample Average Approximation (SAA) method was used to handle uncertainty, converting probabilistic variables into a robust deterministic equivalent that prioritizes electrical proximity during reconfiguration. The model was validated using a composite dataset spanning nearly two years of hourly load and renewable profiles. A total of 600 stochastic scenarios were considered and analyzed to represent an empirical distribution of real-world uncertainty while preserving key temporal correlations. Performance was tested under N-1 and N-2 contingency events, in which one or more microgrids are deactivated, to evaluate system resilience. Results indicate that while a single active MG improves the load factor, it also increases operational instability and objective function variance. Conversely, a three-MG configuration enhances system stability and predictability. Economically, the mesh architecture allows for temporary MG deactivation to reduce maintenance and fuel costs without compromising service. The proposed strategy achieves 100% resilience, ensuring uninterrupted service even under severe constraints.