Power System Transient Vulnerability Assessment And Cascading Outages Forecasting

Blackouts happened recently around the world showed how to deal with large-scale power system blackout so as to ensure the normal operation of power grids has become a clear and urgent research issue. To assess the vulnerability and stability of power systems, as well as to strengthen to analyze and predict cascading trip events which probably exist in power system, providing the theoretical basis for formulating correct counter measures, is an important fundamental work that prevent and reduce the occurrence of vicious large-scale blackouts in power grid.This thesis, starting from the fault may occurred in power grid, launches deep researches on assessing vulnerability of power grid, assessing transient stability of power system and cascading trip events caused by overload and other factors on condition that the power system did not lose transient stability, the main research work are as follows:Contingent faults'probability models taking the weather factor into account are set up, and the idea of evaluating a power system's vulnerability based on the system's transient energy margin is proposed. In order to avoid a serious fault of smaller occurring probability is ignored, the idea of using both probability index and energy margin index to evaluate a system's vulnerability is put forward. The formula to calculate the energy margin index of a fault is defined based on the transient energy function, which can reflect the influence of load demand change to the system's vulnerability. The weak components in the system can be found by means of energy margin indices relating to faults. Moreover, based on probability theory and norm theory, the models to calculate compositive probability index and energy margin index of a power system are derived. The models'parameters are probability indices and energy margin indices of all possible faults in the system. Based on the models, an assessing algorithm for power system vulnerability is proposed.A method of applying Boostting Bayesian classifier to power system transient stability assessment is proposed. In order to avoid collecting the unnecessary attribute variables, the variables reflecting power system transient process are extracted by using grey relational clustering method, and then 12 characteristic variables are selected to be attribute variables of Bayesian classifier, at the same time take stability or instability as class variables. Because the Bayesian classifier can only handle the discrete variables, continual attribute data is dispersed based on information entropy and rough set theory. According to AdaBoost algorithm, single Naive Bayesian classifier is boosting, and then a boosting united Bayesian classifier using to assess transient stability is constructed which reduces the misclassification rate effectively.At the basis of analyzing the cause of cascading failures triggered by line overload, a method of predicting chain trip sequence event is proposed. In order to avoid breaking line blindly and screen the cascading failures mode which causes serious consequences and is prone to happen rapidly, assign the initial breakdown, calculate the power flow using DC power flow model, define the index which reflect the severity of overload, and then a prediction method is proposed based on the severity of line overload considering the impact of hidden failures according to the heuristic search strategy. This method proposes generalized hidden fault model based on the original fault model, and this model forms the hidden failure set taking into account the fault lines caused by hidden failures of relay protection and the hidden failure line which is in the same transmission section as the original fault line. The fault lines caused by hidden failures of relay protection are composed with the lines connected to the both sides of original fault line; the hidden failure lines which are in the same transmission section as the original fault line can be find out by chart searching algorithm after dividing the network topology map into several regions according to graph theory knowledge.