Importance sampling for power system security assessment

A Monte Carlo simulation scheme for estimation power system network failure rates has been developed and implemented. It uses an importance sampling strategy which biases the simulation sample towards the extreme conditions leading to loss of supply. A theoretical analysis of the performance of the sampling scheme is given, and numerical experiments both with an idealised 3-node network and 22-node model of the UK South-West peninsula 400 kV network confirm predictions of large improvements in efficiency, compared with standard Monte Carlo sampling runs using 1000 simulations suffice to determine system failure rates to within 5-10%. Similar gains in efficiency are obtained using a demand disconnection index of system failure, rather than a simple fail/no fail criterion.<
>