Disaster Response Failure Recognition For Power Grid Based On Fault-tree

Natural disasters, man-made disasters and the accidental failures of power grid areall easy to cause power grid damaged and spread to other critical infrastructures, and itsnegative consequences may be multiplied with failure chain effect. It needs to obtainenough useful data and information to support decision analysis for identifying risk. Forthis reason, we can control the situation of the disaster and avoid its expansion underlimited response capacity. Furthermore, responders may face uncertain, fuzzy andstochastic complex context. How to find out the key response failures becomes a hardresearch topic. In this paper, taking disaster response failure as research object, acorresponding theory and method framework are proposed based on fault-tree analysis.First, it describes the disaster situation elements including disaster-causing factor,disaster-affected body, disaster-formative environment and disaster-response body, andresponse failure as driving factor promote the situation evolution that towards tounexpected consequence. The generalized concept of response failure is defined which isother than the traditional failure concept. Then, the source of response failure is tracedfrom three aspects that involves unexpected disaster situation, emergency prepardnessinsufficient and response inappropriate. According to failure effect, it can be divided intotwo categories response-termination failure and capacity-reduction failure and discusstheir effect on rescue in initial and middle stage. At last, the failure sub-systems’conceptual model of “three-stage task” can be constructed.Second, unexpected events can be extracted to form three fault trees of situationcoginition, resource management, temporary power supply and power repair. Moreover,depending on specific situation, the generation processes of situational fault-tree storedby XML for sharing and transfering data are given by using Delphi method.Then, the minimal cut set can be obtained by converting the fault-tree to binarydecision diagram combined preferred method of adjacent nodes. On the other hand, inorder to deal with uncertain environment, the D-S evidence theory and trapezoidal fuzzynumber are introduced into qualitative semantics evaluation method to generate thefailure probability. Then, the probability of top event, the importance degree of eachminimal cut set, the sensitiveness of each bottom event can also be obtained. At last, thegray correlation degree is employed to identified the most key failure set.Finally, a case study is given that we choose Shenzhen as the verified city toillustrate the proposed models and methods for application.