Abstract
We generalize an analytically solvable probabilistic model of cascading failure in which failing components interact with other components by increasing their load and hence their chance of failure. In the generalized model, instead of af ailing component increasing the load of all components, it increases the load of a random sample of the components. The size of the sample describes the extent of component interactions within the system. The generalized model is ap- proximated by a saturating branching process and this leads to a criticality condition for cascading failure propagation that depends on the size of the sample. The criticality con- dition shows how the extent of component interactions con- trols the proximity to catastrophic cascading failure. Im- plications for the complexity of power transmission system design to avoid cascading blackouts are briefly discussed.