Research On Dual-stage-search Algorithm And Risk Assessment Model For Power System Cascading Failure

Power system cascading failures have obtained wide concerns all over the world during the past few decades. With the increasing of grid interaction and correlations between grid components are increasing. Local component failures may result in wide range overloads as well as voltage over-limit problems, which would lead to system-wide cascading failures and large-scale blackouts under extreme condition. Hence, the understanding of the cascading failures, including the conditions of occurring and paths of development, the probability of occurrence and extent of risk to the loads, will play an important role in the operation security of power system.In this thesis, the “8.14” large-scale blackout happened in the USA and Canada on August 14, 2003 is discussed in detail at the beginning and the development process is classified into two stages, i.e. the slow dynamic stage and the fast dynamic stage, according to the differences on the time scale and mechanism of occurrence. Then a new cascading failure probability model is proposed with consideration of the influence of the environment temperature and response delay of the operators.A graph database tool, Neo4 J, is adopted to set up the geographic and lightening database about the main transmission lines in the selected regional power grid. The correlation between the lightening trip and the length of lines in the lightning region is analyzed and used to set up the climate-related initial fault selection model in the cascading fault analysis. A fast cascading-fault-tree-search algorithm that can search different fault developing path in parallel, is designed based on a heuristic function considering both the probability and severity of consequences and by expending the orders of fault sets. This method helps to mitigate the confliction between searching speed and accuracy and reduce the loss of useful information.To deal with the abnormal events during the cascading failure analysis, such as power flow divergence and the emergence of electrical islands, a DC power flow based load curtailment algorithm is developed based on sensitivity calculation and a failure state-matching module is designed to improve the efficiency of load curtailment calculation. The indices and framework for the risk assessment of cascading failures are also introduced.A software based on Java is developed to realize the proposed algorithm in application to practical power systems, where the power flow is calculated by invoking the related PSD-BPA module into program. Applications on the IEEE 39 bus system and the equivalent network of the selected regional power grid of year 2011 are presented as examples to verify the proposed algorithms. Both the cascading fault trees and the cascading failure risk indices are obtained and the main factors to influence the risk of cascading failures are analyzed.