Research On Method And Its Application For User Power Consumption Behavior Mining Based On User Power Consumption Data

The identification of relation between the user behaviors and the power consumption on the user client in time based on the variety of power amount by user can not only make building behavior energy saving in practice,but also can raise the big data based intelligent level of application about security,intelligent district and intelligent community,the fine management of society.In this thesis,the focuses of our research are around to the method of the mining for user behavior on power consumption based on user power consumption data and the application of the method.To mine user behavior on power consumption based on user power consumption data,data estimation method for two adjacent data missing problem and identifications method for the power related event of user are researched and the using of those methods are discussed to.Firstly,to estimate the missing electricity consumption amount in period,the formalized description of the two adjacent data missing problem is given and the value estimate problem for missing item is modeled as one multi-objective optimization problem with the summation as constrains.To evaluate the performance of estimate method for the problem,evaluate criterion for estimate method is also presented in this paper.To solve the problem,one feed forward neural network with some historical items as inputs is used as the estimate function and the connection weight matrix and bias of the neural network are optimized with PSO algorithm.The fitness function of the PSO algorithm is constructed according to constrains based multi-objective optimization problem.To test the performance of model and method presented in this thesis,electricity meter recording data of a residence community is used and experiment results show that the accuracy and stability of our method and methods is high.Secondly,to identify the user power behavior,the power data sequence with specified length is used to describe the power event related to the power behavior of user in this thesis.Method based on K-means clustering algorithm for the mining the power data sequence pattern is designed.And method for the judgment that which pattern class that the power data sequence is belongs to is presented in this thesis too.The mining of the power data sequence pattern for a user is the key step for the identification of power event related to the power behavior of user.Experiment results show that the proposed methods are more stable and effective,whether SSE(Sum of Squares for Error)index or silhouette coefficient is used as critical.Finally,one example for the using of methods presented in our research in this thesis is given.In the example,power data of a microwave oven in Anhui province key laboratory of intelligent building and building energy saving is analyzed.Works for analysis of those microwave oven power data include the process of missing data,the preparation of power data sequence set,the power data sequence pattern for the user and the identification of abnormal behavior related power consumption.With methods presented in this thesis,abnormal behavior related power consumption can be identified.The identification of abnormal behavior related power consumption is one of fundamental for IT application such as security,intelligent district and intelligent community,the fine management of society.With the identification,big data based IT application about security,intelligent district and intelligent community,the fine management of society can be more smarted and the intelligent level of those applications can be raise up significantly.