Improving Monte Carlo simulation efficiency of level-I blackout probabilistic risk assessment

Blackouts in power systems are due to cascading failures whose typical development can be split in two phases: a slow cascade and a fast cascade. Once a blackout occurred, the restoration is as an additional (and last) phase. The blackout Probabilistic Risk Assessment (PRA) can be decomposed in three levels, according to three phases. An analog Monte Carlo simulation has been developed for the level-I, in order to simulate independent and thermal failures during the slow cascade. The main limitation of such an analog simulation is the small fraction of runs leading to interesting consequences. The aim of this paper is then to propose biasing techniques in order to improve the blackout PRA level-I Monte Carlo simulation efficiency. Two methods are explored: favoring failures during the cascade by forcing them to occur before a time limit and favoring thermal failures by biasing weather conditions sampling. Results obtained on a test case show that a significant gain can be reached.