A Scalable Risk-Aware Loop-Based Matheuristic for Reliability and Resilience Enhancement in Active Distribution Networks

Abstract

Active distribution networks require scalable reconfiguration strategies capable of improving reliability and resilience under renewable generation variability, demand uncertainty, and component outages. However, extensive-form mixed-integer formulations rapidly become computationally challenging as the number of scenarios and candidate switching actions increases. This paper proposes a scalable risk-aware loop-based matheuristic for reliability and resilience enhancement in active distribution networks. The method exploits the fundamental-loop structure created by normally open tie-lines to encode the reconfiguration problem as the selection of one open branch per loop, thereby reducing the topological search space while preserving radiality and connectivity. A multi-start population-based search with local improvement generates high-quality radial candidate topologies. Each candidate topology is then assessed through a continuous scenario-based operational optimization model that considers active and reactive power balance, linearized voltage constraints, renewable utilization and curtailment, battery energy storage support, load curtailment, and critical-load restoration. A conditional value-at-risk term is incorporated to penalize severe energy-not-supplied outcomes and improve protection against tail-risk scenarios. The framework reports conventional reliability indices, including EENS, AENS, SAIFI, SAIDI, CAIDI, and ASAI, together with a resilience index based on weighted critical-load service. Numerical studies on the IEEE 33-bus, IEEE 69-bus, and IEEE 123-bus distribution systems demonstrate the performance of the proposed method over feeders with increasing size and switching complexity. Additional nonlinear AC power-flow validation, out-of-sample uncertainty testing, and optimization-baseline comparisons are used to assess the physical feasibility, generalization capability, solution quality, and computational behavior of the proposed framework.