Coordinated Restoration of Integrated Electricity-Gas Systems With Frequency Stability Constraint Learning: A Multi-stage Model with Exogenous-Endogenous Uncertainties

The ability of integrated gas-electricity distribution systems (IGEDSs) to survive under extreme disasters attaches great importance. This paper proposes a multi-stage coordinated restoration model of IGEDS considering exogenous-endogenous uncertainties with initial and conditional nonanticipativity conditions under disasters. The exogenous uncertainties consist of component failure uncertainty and energy generation/ consumption uncertainty, and endogenous uncertainty represents traffic uncertainty determined by repair crew routing decisions. Network reconfiguration and frequency reserve of distributed generators (DGs)/ wind turbines (WTs) are utilized for enhancing system resilience. The optimization with constraint learning (OCL) method is applied to build linearized frequency nadir constraints via dynamic sparse neural network training and pruning. Moreover, to solve the proposed model with exogenous-endogenous uncertainties, an enhanced outer approximation (EOA) algorithm is presented. Numerical analysis indicates that the proposed model could effectively improve the load restoration of IGEDS under extreme disasters.