Research On Estimation Methods For Cascading Failures In Power System

With the interconnection of power grids contributing strong support to economic in our country,new challenges have appeared in the safety and stability operation of power system.In recent years,frequent large-scale blackouts happening worldwide have beenalarming to experts in power system.Researches around the world show that,large-scale blackouts happen mostly in relation to cascading failures.Thus,the research on power system cascading failure is still a major and urgent topic.The main research methods on cascading failures are complex network theory,complex system theory and power system simulation method.Since the research on power system cascading failure encounters a great amount of calculation,this dissertation studies how to speed up calculation in cascading failure situation,and obtains the following results:1.Based on complex network theory,this dissertation proposes a comprehensive framework for the vulnerability analysis of cascading failure to speed up the computation.Firstly,the traditional topology based graph model is improved by depicting a power grid as a weighted graph based on the reactance matrix.Secondly,the concept of load is redefined using power angle information.Thirdly,the power flow constraints are adopted to speed up the computation instead of the shortest path based flow scheme.Simulation results also demonstrate that the proposed one is more effective and efficient for the vulnerability analysis of power system cascading failure.The CDF curves of the proposed model obey a power-law distribution.2.For the tremendous amount of power flow calcution,this dissertation proposes an efficient power flow calculation method of cascading failure.A new index is proposed to monitor the error and indicate when to calculate AC power flow.Based on the distribution factors and the new index,a fast estimation algorithm is proposed to balance accuracy and efficiency in cascading failure analysis.The simulation results show that the computation efficiency is greatly enhanced and computation accuracy is guaranteed.3.For the tremendous following-up failures scenarios,this dissertation proposes a risk based identification methodology of cascading chains to select the cascading tree.Based on distribution factors and pre-contingency power flow conditions,a novel risk index is proposed to quickly assess the consequences of possible cascading failures.Hence,compact cascading tree and key cascading failure sequences can be determined and ranked according to the risk index.The simulation results show that the proposed risk assessment method can reduce cascading analysis computation time.4.For the transient control cost in the cascading failure analysis,this dissertation addresses the necessity of cascading analysis should develop from steady state analysis to transient analysis,then studies the fast calculaton of control cost of transient frequency fault based on Extreme Learning Machine(ELM).This dissertation proposed a methodology for real-time predicting required FDLS:the general idea is to train an ELM-based prediction model with a strategically prepared FDLS database,and apply it to on-line for real-time FDLS prediction.The methodology can speed up the calculation greatly which is promising for practical use in the cascading risk analysis in the future.