We describe a model of power networks which may be useful for studying power blackouts. The model is a combination of the admittance model of the network and the probabilistic model of faults of its components. To compute the probability of a power outage possible failure events are considered and in each case the possibility of a cascading failure is studied. Graph representation of the network is used to detect connected components in the network, and then network equations are solved separately in each component. We show how to use this model to compute the probability of a power blackout in power networks and how this model can be used to develop suggestions for improving power network designs. The IEEE 188 bus is considered to show the usefulness of this approach.