Multiple Resources Information Fusion For Power System Fault Diagnosis Based On Improved Bayesian Network And Hilbert-Huang Transform

With the continual enlargement of power system scale, when faults occur in power system, a large number of alarms are sent to the dispatch center, it requires dispatcher judge the faulty component quickly and accurately. Therefore, developing a effective fault diagnosis system is very significant for guaranteeing power system operation safety and stability. In the case of some complex fault situations, the diagnosis result may be not accurate by only use digital information, However, multiple data information like digital and electrical data is required for fault diagnosis can improve the accuracy of diagnosis results. So, a kind of fault diagnosis method based on multiple data resources is studied in this paper, the improved Bayesian network and Hilbert-Huang transform are used to extract the fault feature of digital information and electrical information and determine the faulty component by data fusion.Firstly, a kind of structure-variable Bayesian network power system fault diagnosis considering credibility was studied in this paper. the existing Bayesian network power system fault diagnosis is influenced by complicated structure of power system and the uncertainty of protective relays and circuit breakers information, it results in complicated network model and incorrect diagnosis results. According to judge the fault patterns based on related protective relays and circuit breakers information to construct corresponding Bayesian network models. Then, the credibility of protective relays and circuit breakers action events of Bayesian model are estimated and taken into the improved formula of bayesian probability, calculating the failure probability of component. The proposed Baysian model is simple and it can deal with the uncertain of some information effectively, simulation cases have illustrated that the proposed method is effective.Secondly, a kind of fault diagnosis method based on multiple resources information fusion was studied in this paper. Gathering the direction protection and distance protection zone I and zone II information of component in power supply interrupted region to construct fault judgement matrix, according to to reduce the number of suspicious faulty component according to the judgement of suspicious faulty component method, transfer the current record data of suspicious faulty component, analyze the changes of current amplitude and energy by using Hilbert-Huang transform method to get the distortion degree of current amplitude and the value of faulty energy, the improved D-S evidence theory is used to fusion the relative fault degree of current amplitude and energy with the fault degree of Bayes network relative fault degree, determine the faulty component through the result of data fusion, some simulation cases have illustrated that this method can judge the faulty component under complex fault situations effectively.