Vulnerability Assessment Of Power Grid Under Extreme Events Based On Graph Embedding Theory

With the development of technology,electricity has brought great convenience to human beings.Power systems which realize the production,transmission and distribution of electricity is becoming the most important infrastructure of human beings.However,in recent years,extreme events,such as natural disasters and deliberate cyber attacks,pose a serious threat to the power grid and bring great challenges to its reliability.In addition,the power system is a complex system connected with a large number of components.A single failure of some vulnerable components may lead to the failure of other components,so that the failure spreads and evolves into a cascading failure,leading to large-scale power outage.Therefore,it is urgent to assess the vulnerability of a power grid to extreme events in order to reduce these negative impacts.In order to cope with this problem,it is necessary to study impacts of extreme events on power grids,and take corresponding measures to assess the vulnerability of power grids.Aiming at this subject,this thesis investigates several extreme events and researches on the vulnerability assessment of power grids under such events,mainly including the following four aspects:(1)This thesis studies the feasibility of using graph embedding theory in the vulnerability identification of a power grid.In this thesis,we introduce the structure models and characteristic indicators commonly used in complex network theory,and attempt to describe the complex connection of a power grid as the connection of nodes and edges by the complex network theory.Therefore,it is necessary to study the impact of extreme events on the power grid and identify the vulnerability of the grid to such events,to reduce the negative impact.(2)This thesis develops a severe false data injection attack(FDIA)identification method based on Deep Walk algorithm.Aiming at studying the scenario that hackers launch FDIA attacks on a power grid,a two-layer optimization model is applied to constructed FDIA attack vectors from the attacker’s perspective.Based on the analysis of cycber attack mechanism,the response model of the power grid under FDIA is established.After obtaining abundant cycber attack scenarios,the association graph of attack scenarios is constructed based on batch random edge reduction strategy.Then,the graph is set as the input of Deep Walk algorithm to learn low-dimensional vectors of nodes in the graph,so as to realize the identification task of severe FDIA attacks.The performance of the proposed method is evaluated by FDIAs simulations in the IEEE 30-and 118-bus systems.(3)This thesis attempts to build a double-layer graph embedding algorithm to identify the vulnerable transmission lines in a transmission grid.The structure and attribute network embedding(SANE)is proposed to learn the low-dimensional vector representations of lines in the grid.Firstly,the grid is represented as a two-layer network which preserves topological and electrical properties of transmission lines.Secondly,the partial random walk strategy of virtual particles in the two-layer network is defined,including intra-layer and inter-layer walk.Then,the low-dimensional vector representations of transmission lines is learned based on the obtained random walk sequences.Finally,the vulnerability assessment index of transmission lines is proposed according to the low-dimensional vectors,and the calculation examples are verified in the case＿ieee＿rts＿2＿area and nesta＿case118＿ieee systems.(4)This thesis proposes a SANE based method for identifying vulnerable transmission lines under a hurricane and a resilience assessment method of a grid.Firstly,the response model of a transmission grid under a hurricane disaster is established based on analyzing of the failure of lines.Then,the cascading failure model is constructed according to the response model,and the cascading failure graph under persistent disturbances(CFGPD)is set as the input of SANE to identify the vulnerable transmission lines.In addition,a resilience assessment index of a transmission system under extreme weather events is proposed to evaluate the effectiveness of the resilience enhancement measures.Finally,the proposed methods are demonstrated on the Texas power grid under Hurricane Harvey.The performance of a power grid under extreme events is studied,and the vulnerability of the power grid is identified based on the graph embedding theory.We have compared many state-of-the-art methods on several cases and the results prove the effectiveness of the proposed identifying methods.This thesis provides supports for improving the resistance ability of a power grid under extreme events.