Probabilistic resilience assessment of urban distribution power grids by fast inference of multi-source multi-terminal network reliability

Abstract

Urban power distribution grids featuring loopy topologies and integrated distributed generations pose significant challenges for efficient and precise resilience quantification against disruptive events. This paper presents a probabilistic resilience assessment framework tailored for such grids. Risk metrics grounded in loss of load probability (LOLP) and expected energy not served (EENS) are formulated to evaluate resilience across multiple temporal stages. A multi-source multi-terminal network reliability (MSMT-NR) modeling approach is proposed to characterize the stochastic impact of component failures on load point connectivity. A computationally efficient algorithm framework is developed for the inference of the MSMT-NR problem, comprising: (1) Derivation of analytical LOLP expressions for grid topologies exhibiting tree-like load subgraphs; (2) A deletion–contraction decomposition technique generating solvable tree subgraphs from arbitrary network structures; (3) A computational graph-based inference methodology enabling efficient MSMT-NR evaluation and automatic differentiation for sensitivity analysis of component importance measures. Strategies for enhancing scalability to large-scale grids are devised. Extensive case studies on a real-world 30,894-node distribution grid corroborate the efficiency and precision of the proposed approach.