Research On Key Technologies Of Big Data Analysis For Distribution Network

The application of big data technology in distribution network includes noise data identification,distribution network planning,load forecasting,fault diagnosis,intelligent distribution and so on.Therefore,the application of big data technology in the operation of distribution network can not only meet the user’s demand for electricity,but also save electricity.According to the historical data collected by the power grid company,this thesis mainly studies the noise data recognition,power consumption behavior analysis and load forecasting in the distribution network environment.By consulting the data,we can see that it is difficult to achieve good results by using only one big data technology.The traditional k-means clustering needs to determine the number of clusters in advance and randomly select the initial clusteri ng center.This thesis proposes k-means clustering based on correlation.In the aspect of electricity usage behavior analysis,the traditional clustering and classification methods are ineffective in identifying electricity theft users.It is difficult to accurately classify the electricity usage behavior of user s.A complementary analysis method based on k-means and DBSCAN is put forward in this thesis.The explicit contents of this substance are as follows:Firstly,this thesis uses k-means clustering based on correlation,which can effectively divide the highly correlated data into a cluster,and verifies the effectiveness of the proposed algorithm by clustering indicators.The processed data sets are identified by decision tree method in the second stage of noise data recognition.AUC is used to verify the accuracy of classification,and the identified noise data are removed and picked up.KNN substitution method based on distance distribution weight is used to replace noise data.Secondly,this thesis uses the complementary method based on k-means and DBSCAN to analyze power consumption behavior.Traditional k-means algorithm classifies all data into clustering,even if the correlation is not strong,so it is difficult to identify outliers.This thesis combines DBSCAN algorithm with k-means,which can divide data clusters according to power consumption behavior,so as to distribute power reasonably.Finally,this thesis combines clustering algorithm,uses Elman neural network for load forecasting,uses smoothing function and clustering method to clean data,and verifies the effectiveness of the proposed method through RMSE and related indicators.