Research On Specific Scene Non-Intrusive Load Identification And Fault Arc Detection Technology

In recent years,more and more attention has been paid to safe and intelligent electricity consumption in China,and the research of non-intrusive load identification and fault arc detection technology has been widely concerned.At present,most of the research on non-intrusive load identification technology focuses on ordinary residents,and little attention is paid to the specificity of load in specific scenes,which leads to the practical application effect of specific scenes cannot be guaranteed.To solve this problem,this thesis proposes a non-intrusive load identification and fault arc detection method based on specific scenes.Different from non-intrusive load identification in common home scenarios,dormitory,shopping malls and other specific scenarios are more directional,practical and feasible due to management needs.In this thesis,the specific scene of college dormitory is taken as an example to study.Firstly,the common electricity load of the dormitory is classified,and the load and characteristics are analyzed according to the electricity consumption rules of the scene,and four new characteristic quantities are proposed to be applied in the followup work.Aiming at the problem of load switching error detection caused by power fluctuation and complex load operating state,this thesis proposes a load switching event detection method based on power fluctuation and sliding window Chi-square goodness of fit.The measured data and Plaid data set were used to simulate the dormitory scene and the detection method was verified.At the same time,considering the concealment and high harmlessness of series fault arc,this thesis proposes a series fault arc detection method based on dimensionless characteristic indexes,and designs a fault arc experimental platform for collecting actual arc data and verifying the method.The accuracy of offline data and hardware real-time arc detection is 98.0% and 95.0%.On the other hand,in the load identification process,Fisher linear judgment is used to analyze the importance of features according to the load power and establish the feature database.At the same time the use of unsupervised fuzzy c-means clustering and supervise two support vector machine(SVM)algorithm based on scene dormitory load identification verification,validation results show that the two methods of single load and load the recognition accuracy of 91.0% and 89.0%,so the method proposed in this thesis the scene in the dormitory of load identification and has good application value in the arc fault detection.