Identification of Cascading Failure Propagation Under Extreme Weather Conditions

As a fundamental infrastructure, power systems play a vital role in modern society, but it can be damaged by different adverse events e.g. natural, accidental, and malicious, of which the adverse natural events, especially extreme weathers, with huge destructive force can bring tremendous damages and economic losses. The high exposure and comprehensive geographical coverage of the power system make it highly vulnerable to extreme weathers, resulting in equipment damage which leads to cascading failures and blackouts. Traditional methods only focus on modeling and analysing the reliability of the power system under extreme weathers, without focusing on the propagation of the cascades. In this thesis, innovative methods of studying the cascading failure were proposed, and further extend to collectively consider the impact of extreme weathers on the transmission networks. The proposed models were further validated by applying them to a study system (IEEE-30 bus system) and a real system (Italian transmission network). A so called normal failure model based on probabilistic graphs was proposed to describe how a cascading failure propagates under a contingency analysis. This model employed Monte Carlo simulation to consider most of the possible operating conditions to establish directed probabilistic graphs to identify the cascading propa-gation by tripping all branches one by one under each operating condition. Obviously, the results of the model can clearly and legibly show the main cascading path of a given network without considering the initial operating condition and the triggering contingency. Further, an index based on branch vulnerability was designed to select the triggering event to increase the effectiveness of the failure in the simulation. Furthermore, by integrating a probabilistic model of extreme weather impact into the normal failure model, the extreme weather model was proposed based on failure networks, which maps a physical electricity network into a graph in the cascading propagation dimensions. Based on the generated failure networks, a new method based on clustering techniques was proposed to fast track the cascading failure path from any initial contingencies without recalculating the cascading failure in the physical network. The high similarity of the simulation results on the IEEE 30 bus system from the two proposed models indicates the validity of the models. Further, to demonstrate the extreme weather model, we selected a winter storm, which could happen in Northwest of Italy as an example. The data of snowfall on the Alps was collected and modeled by probability density function and probability mass function. By applying the proposed extreme weather model, the propagation paths can be predicted. The values of the study provide two powerful tools which can 1) clearly present the inherent characteristic of any one given network, i.e. main propagation paths exist regardless of the initial network and failure condition; 2) fast and reasonably predict the cascading paths in a network under extreme weather conditions.