Statistical Properties and Classification of N-2 Contingencies in Large Scale Power Grids

We present the systematic analysis of all dangerous N - 2 contingencies observed in medium size model of Polish power grid with about 2600 power lines. Each of the dangerous contingencies is composed of two initially tripped lines and one or more lines that overloaded as the result. There are 443 distinct contingencies that do not lead to immediate islanding of the grid. In the scope of the work we analyze the statistics of individual line participation in those contingencies and show that some lines have anomalously high rate of participation in the contingencies. Next, we show that about third of all the contingencies can be associated with the sub grids that are connected to the rest of the grid via small set of power line chains. The contingencies arise when cutting some of those chains results in overload of the others. Simple reduction of power grid corresponding to aggregation of chain components significantly reduces the total number of distinct contingencies. The rest of the contingencies are closely related to a set of almost dangerous N-1 contingencies that result in heavy loading of particular lines. Tripping many different additional lines on top of these N-1 contingencies results in an overload of one or more lines. We conclude our work by characterization of the joint distributions of power flows through the initiating and overloaded lines and statistical analysis of topological distance between the initially tripped and overloaded lines.