Research On K-Means Clustering Algorithm For Residential Electricity Data In Energy Internet

The development of Internet technology and the high requirements for power have expedited Energy Internet.In this context,residential electricity usage is gradually intelligent,and achieving the electricity information two-way interactive between the Grid and the users has become an inevitable trend.Meanwhile,with the continuous development of intelligent community,a large amount of electricity consumption data is accumulated.Through mining these information by data mining methods and analyzing the users' electricity consumption behavior,it can provide data support for the Grid companies formulating scientific and reasonable response strategy of power demand.Moreover,these useful information can be passed to residents,helping the users understand themselves' electricity usage situation,regulating electricity consumption behavior and taping the potential of electricity usage.Based on the current research of domestic and foreign scholars,the research on K-Means clustering algorithm for residential electricity data in Energy Internet is carried out,and the main research work is as follows:1.In order to solve the problem of determining the initial clustering center and value K in traditional K-Means algorithm,an improved K-Means algorithm based on density Canopy is researched.Defining the product of sample density,the reciprocal of the average distance between the samples in the cluster,and the distance between the clusters as weight product,the initial seeds are determined by the maximum weight product in the datasets.The density Canopy is used as the preprocessing procedure of K-Means and its result is used as the clustering number and initial clustering center of K-Means algorithm.The simulation results show that the improved algorithm achieves better clustering results and is insensitive to noisy data.2.A data analysis method of power utilization based on improved K-Means algorithm and cloud computing is designed.For the electricity data in smart community,the parallelization of K-Means algorithm based on MapReduce model is carried out,mining its effective information.Taking family as unit,calculate such features as electricity consumption rate during peak hour,load rate,valley load coefficient and the percentage of power usage during normal hours,and establish the data vector dimension,thus completing the clustering of similar user types,while analyzing the behavioral characteristics for each types of users.Experiment clustering results shows that the method can be stable and efficient operated and it can achieve better clustering effect.3.The short-term electricity load prediction approach of BP neural network based on K-Means algorithm is designed.Firstly,the clustering analysis for users' historical electricity consumption data based on the improved K-Means algorithm is performed.Then,based on the clustering results,selecting the historical data for the type of the user to be predicted and the work of short-term electricity load prediction is launched.Finally,the contrast experiment indicates that the BP neural network prediction method with improved K-Means clustering analysis can achieve better accuracy than that of the method which has not used K-Means clustering analysis method.Finally,the work of the dissertation is summarized and the next step of the research is discussed.