Research On Smart Grid User Side Energy Consumption Data Monitoring System

With the rapid development of data mining and artificial intelligence technology,utilization of the decomposed data of family load and large amounts of date that are generated during the using of appliances has important values for the improvement of energy-saving and the development of smart grid.In big data Era,data analysis requires that the energy measurement system can decompose the appliances electricity data in one power unit,but the traditional method can only measure the total energy consumption of each power user.The traditional energy measurement method has not been able to meet the requirements of the data decomposition.In this paper,the design of smart grid user side energy consumption data monitoring system can realize the decomposition of household load electricity data.Users also can use the system to get their detailed power consumption information.Currently,compared with intrusive load monitoring,non-intrusive load monitoring is the main direction in the load monitoring field because of its lower cost.However,this model also has certain limitations.First of all,this method is not practical for family load monitoring because the building of hardware platform needs to transform large-scale home power line.Secondly,in order to guarantee the accuracy of computation,computing load that involved is very large.To solve such problems,this paper designed a novel system structure for load monitoring.The system structure incorporates the structural advantage of intrusive load monitoring and non-intrusive load monitoring.In order to remove the mixed samples,this paper makes an improvement on the repeated editing nearest-neighbor method based on the characteristics of the monitored subjects and the characteristics of classification.By utilizing the distance between sample and clustering centers which is generated in the process of C mean classification.The improved method reduced the editing range of the repeated editing nearest-neighbormethod.Thus,the computation was reduced.Considering the aging phenomenon of appliances,this paper only save the sample which generated during the last 45 minutes of system operation.Other samples will be deleted.The database will be updated every 15 minutes.In order to verify the effectiveness of the designed system in this paper,we selected9 kinds of household appliances that were connected to 3 nodes for the identification experiment.During two-hour experiment,the accuracy of identification can reach99.98%.