Quantifying failure propagation in electric power transmission systems

Understanding large blackouts and quantifying their risk is important because of their large impact on our society and their many consequences for the design, operation and regulation of the electric power grid. Many large blackouts start with small initial failures and propagate in a cascade of further failures. This thesis quantifies the propagation of cascading failures in the power transmission network.;Bulk probabilistic models of cascading failure include Galton-Watson branching processes. This thesis extends these models in several ways. While the Galton-Watson process is for discretized data, a method is suggested to generalize it to continuous blackout data such as load shed. This work also analyzes cases in which the amount of propagation varies with the number of initial failures or varies as the cascade proceeds and suggests new models.;The splitting method that can greatly increase the speed of simulation of the failure propagation is explained and tested. The key idea of the splitting method is saving the simulation state at stages and reloading and resampling them to continue the simulation. By using the splitting method, it is possible to detect possible clustering in the power network that seems to arise from barriers which tend to inhibit the propagation of failures in the network.;Finally, a method to determine the relation between several indicators of blackouts is explored so that one can estimate the distribution of failures better.