Fault Line Selection Method For Single Line To Earth Fault In Non-effectively Grounded System Based On Big Data Research

Most distribution networks in China are non-effectively grounded systems,which include unearthed neutral system,Peterson coil compensated neutral system and high resistance-grounded neutral system.Neutral indirect grounding can improve the power distribution reliability because the system continues power supply for a while after single-phase-to-earth fault occurs,while,this kind of operating mode results in that it’s difficult to select fault lines because of small fault current,complex fault conditions and unstable arc.Multiple kinds of fault line selection methods are proposed,but the accuracy can’t meet the actual needs,because of this,a large number of substations select fault lines by open feeders in sequence until the fault disappears.With the development of social economy,our requirement of power supply reliability and power quality has raised,researching a high-accuracy fault line selection method to improve power supply reliability,customer satisfaction and reduce outage costs is significant meaningful.On the basis of existing fault line selection methods,this paper studies the fault line selection method for non-effectively grounded system based on big data.First,collecting fault-related data,preprocessing and feature extraction are conduct to convert raw data to feature data set.Next,data mining algorithms are applied to build the fault line selection model,feature data set is applied to train the model,and the well trained model can be used to select fault lines when new fault occurs as criterion.Next,when new fault occurs,collect real time fault data and process raw real time fault data to real time fault feature vector,input the fault feature vector into the model and output the predicted value,by evaluating the predicted value,the fault lines can be obtained.Finally,simulation of real non-effectively grounded system is conducted to prove the feasibility of the method proposed in this paper.The main researches are as follows:(1)The existing fault line selection methods are summarized,and the advantages and disadvantages of each method are pointed out.Also,the problems of existing methods are summarized,and the development trend of fault line selection is analyzed.The big data technology and big data platform are introduced,and the application situation of big data technology in power industry is summarized.(2)The fault features when single-phase-to-earth fault occurs in neutral ungrounded system and Peterson coil compensated neutral system are analyzed.The steady state features include the fundamental harmonic steady state fault feature,the fifth harmonic fault feature,the active component fault feature and the feeder fault feature.The transient state features include the first half wave feature,the frequency domain fault feature and the wavelet packet fault feature.Finally,the fault features extraction methods are proposed according to the fault features analysis results.(3)The fault line selection method based big data is proposed.Above all,the’offline training and online application’ framework is proposed,and the fault line selection process is illustrated.On this basis,the key process are explained in detail:first,the data types and data sources are listed;second,the preprocess and feature extraction methods are proposed;finally,the data mining model and algorithms are introduced.(4)Simulation of real non-effectively grounded system is conducted to prove the feasibility of the method proposed in this paper.Software PSCAD is used to generate fault big data,and software MATLAB is used to process the raw data and convert these data to feature data set.Finally,software SPSS Modeler are used to mining the feature data set,and apply the trained model to select fault lines.Finally,the result proves that the method proposed in this paper is feasible and improve the accuracy of fault line selection.(5)Summarizes the full text of this paper and prospects the work in the later stage.