A Reliability Assessment Method Integrating CGAN-Generated Source-Network-Load Scenarios for Active Distribution Systems

Carlo simulation (MCS), often fail to capture complex, high-dimensional stochastic interactions between source-network-load (SNL) uncertainties. This paper proposes a novel reliability assessment framework integrating Conditional Generative Adversarial Networks

(CGAN) with MCS.

To address these limitations, this paper proposes a novel reliability assessment framework for

ADSs that integrates CGAN with MCS. The CGAN-based framework generates temporally coherent supply and demand trajectories that preserve spatial correlations among load nodes and distributed generator (DG) outputs. These generated scenarios are paired with sequential system-state realizations (including fault events) sampled by MCS. Subsequently, a failure consequence analysis evaluates connectivity and the impact of outages on load nodes. To enhance system resilience, a DG-supported restoration optimization model is implemented to identify microgrid boundaries and maximize critical-load recovery. Finally, reliability is quantified through condition-specific indices, accounting for varied demand scenarios, failure consequences, and DG-supported recovery effects. The methodology was validated using real-world data from the Northwest Power Grid of China and the IEEE 37-bus system. Results confirm the proposed framework accurately captures distinct reliability variations across different conditional scenarios, providing a high-resolution tool for future grid planning.