| dc.description.abstract | Modern power systems are immensely complex, and the reliability of electricity supply is
governed by rare interruption events which sometimes invoke catastrophic consequences.
Proper reliability analysis is crucial both in the operating and planning phase of power
systems, and the risk of interruptions must be evaluated with probabilistic models. Monte
Carlo simulation is widely used in modern reliability analysis, but due to the rareness
of interruptions, a naive Monte Carlo simulation is usually an inefficient approach. This
thesis explores how importance sampling can increase the precision of different Monte
Carlo models. The research culminates in a novel method that combines the principles
of resampling, importance sampling and the cross entropy algorithm. The method is
applicable to time-sequential simulations and requires very few model assumptions. When
applied to a reliable grid configuration where interruptions occur in about 2 · 10−5 of the
samples, an average improvement in precision of 92.8% of the expected energy not supplied
was observed. | |
