A Generic Bayesian Method for Faulted Section Identification in Distribution Systems Against Wind-Induced Extreme Events.

The response time of service restoration for distribution systems (DSs) experiencing a major outage is subject to the progress of fault location. Aimed at expediting fault location and reducing the outage duration under wind disasters, a systematic model to identify the faulted sections for DSs is proposed. To cope with the limited network observability and frequently received incorrect data, information from fault indicators (FIs), protective devices and component fragility analysis are jointly considered. The Bayes’ theorem is applied to evaluate the degree of hypothetical fault scenarios supported by observed data and identify the scenarios with high probabilities. First, logical relations correlating fault scenarios to expected FI overcurrent notifications and protective relay tripping signals are formulated. To fully exploit the information from protective devices, the action sequence of relays with different tripping characteristics are analyzed and unobservable devices such as fuses are considered. Next, the probability to obtain the observed data from the expected ones is derived considering overall uncertainties caused by abnormalities in FI detection and communication, and relay operation. The posterior probability maximization is transformed into a mixed-integer linear programming (MILP) problem. Case studies validate the model effect on improving fault diagnosis accuracy.