Automated Design of Realistic Contingencies for Big Data Generation

The work proposes an algorithm for big data generation based on realistic selection of a set of contingencies for power systems described by undirected graphs. Every contingency is created by eliminating a certain number of elements in the system represented by graph edges. The number of elements as well as the distance between elements of the contingency is randomly selected according to a geometric probability distributions based on historical data. The duration of a fault that starts the contingency as well as the time intervals between elements of the contingency are chosen by sampling from a gamma distribution. In addition, the absence of islands in the system is assessed by analyzing the connectedness of the graph with deleted edges, which is quantified by computing the number of zero eigenvalues of the Laplacian matrix of the resulting graphs. The algorithm is validated on the Nordic 44-bus power system.