Research On Intelligent Diagnosis Method Of Power System Fault Components Based On Recording Data

Although the development of power technology has improved the reliability of power systems,grid fault is still inevitable.After the grid fault,the dispatch center will receive various fault alarms,the fragmented information causes the dispatcher to arrange the appropriate scheduling strategy.So,after the grid fault,it is very important to determine the faulty component using accurate and reliable fault diagnosis methods,and report to the dispatcher,to ensure that the grid resume operation as soon as possible.At present,switching fault alarm information is widely used in grid fault diagnosis.It will determine the faulty component when the protection is correct and the alarm message is not missing.However,the protection relay and the breaker often fail to move and malfunction,which causes many difficulties to the fault diagnosis method,and will give error results.The fault recorder is widely distributed in the transmission system,which can record the complete electrical quantity change information before and after the system disturbance,and provides a lot of accurate and reliable data for fault analysis.This paper focus on the fault diagnosis of power grid based on fault recorder data,and uses the recorded data before the action of the protection system to analyze the grid fault.A fault diagnosis method for the dispatcher is proposed.This method is unaffected by incorrect operation of the breakers and relays,it can give results no matter what happens will other diagnostic methods cannot.First,the article analyzes the characteristics of recorded data and multiple fault features are extracted from the recorder data in time series form.And then,the preliminary diagnostic methods and comprehensive diagnostic methods are proposed based on fault characteristics.In the preliminary diagnosis method,the clustering models of the bus,the line and the transformer are established by using the fault characteristics of each component.After the fault occurs,recorded data of grid components from wide area are extracted,and these data is used for cluster analysis.This clustering method classifies grid components into fault and normal categories,and calculates the credibility of each component as fault class.The analysis of the example shows that the method can quickly and accurately identify the fault components.Fault credibility of preliminary fault diagnosis is a probability measure,and uncertain diagnostic results will be trouble to dispatcher in actual operation.The integrated fault diagnosis method establishes a fault diagnosis analytical model using the relationship between the logic state of the recorder data features and the operating state of the grid component,and will give a definitive diagnosis result.The first step is to determine the solution range based on the preliminary diagnosis result and the fault recorder data collected by the dispatcher.And the topology connection relationship of the suspected fault components is determined according to the association matrix.The second step is to extract the fault features of the components in the suspected fault area,and calculate the state of the actual recorded wave feature quantity of each component according to the set value.And convert the features to logical values based on threshold values.The third step is to assume the state information of each component in the area to calculate the expected fault feature state,and the fault diagnosis model is built by calculating the residuals of expected fault feature and the real fault feature.Since the fault feature is calculated using recorder data before the protection action,the method is not affected by the protection action,and the diagnosis model can accurately determine the faulty component.Finally,the actual grid fault case and simulation examples verify the accuracy and feasibility of the proposed method.