An efficient Monte-Carlo method for power grid reliability estimation

Estimating the failure probability of high-dimensional systems is crucial for infrastructure reliability assessment, especially when failures are rare but have significant consequences, as is the case with power systems.While subset simulation (SuS) is widely used for efficient rare-event probability estimation, its accuracy can suffer from large errors when the failure domain of the system is fragmented or its state space is rugged as in networked systems.This paper introduces skipping subset simulation (Skipping-SuS), an enhancement to the standard SuS that incorporates a skipping mechanism to improve global exploration of the state space.Skipping-SuS increases the transition probability between spatially separated failure regions, thereby increasing sample diversity and reducing both bias and variance in reliability estimates.More generally, Skipping-SuS features reduced error estimation losses to better support decision making.A case study on power grid reliability assessment using IEEE benchmark networks demonstrates that Skipping-SuS outperforms standard SuS by consistently identifying more failure regions and producing better estimates.Although the case study focuses on power systems, the proposed method is general and applicable to a broad range of rare-event problems in high-dimensional settings.