Topology Analysis And Estimation Of Statistical Line Loss For Feeders Based On Big Data Of Power Distribution And Utilization

With the construction and development of smart power distribution and utilization system,utilities have accumulated enormous electric data,which lays a foundation for the application of the big data of power distribution and utilization.Utilizing the big data can not only improve the level of planning,design,and operation of a distribution system,but also have potential value in exploring the inherent regularity of the system and providing much more abundant power services.This paper studies the topology analysis and the estimation of statistical line loss for distribution feeders,using the power distribution and utilization data of Pudong provided by Shanghai electric power company.In order to extract a single feeder topology from the huge amount of topological file in the power distribution system,a topology analysis method of distribution feeders based on CIM/XML topological file is proposed.First of all,regular expressions are used to match useful information for topology analysis from enormous data in the topological file.After that,the association of equipments and nodes is constructed based on the information extracted from the topological file.Then the association is converted into a node information table and a linear devices information table.The depth-first search algorithm is used to resolve the topological information of a feeder.Eventually,the method is verified by the topological file of Lujiazui.Aiming at the solution of lost data and abnormal data problem of statistical line loss of distribution feeder in smart power distribution and utilization system,a novel estimation method of statistical line loss of distribution feeders based on cluster analysis and eXtreme Gradient Boosting(XGBOOST)is proposed.Firstly,a feeder clustering index selection method is discussed,considering the correlation and validity of influencing factors for statistical line loss of distribution feeders.Then the Partitioning Around Medoids(PAM)algorithm with weighting distance is proposed to cluster the sample data of distribution feeders.Secondly,the statistical line loss estimation model of distribution feeders based on XGBOOST is constructed.Then the internal relations between statistical line loss rate and characteristic parameters of feeders including its theoretical line loss rate is mined by training the model with sample data of each cluster respectively.Finally,the method is verified by 1174 distribution feeders in Shanghai-Pudong distribution system.The results show that the XGBOOST method for the estimation of statistical line loss of distribution feeder has higher accuracy than traditional linear regression,decision tree and neural network.