Study On Information Value Added Method Of Electric Power User Side Based On Multi-Source Data Mining

With the development of the smart power distribution and consumption grid,the number of acquisition terminals and the frequency of the acquisition is increasing constantly,which form multi-source smart power consumption big data gradually.The traditional data mining method can't apply to the power consumption big data.It is necessary to research new data mining analysis method which is suitable for the power consumption big data.Through mining the multi-source power consumption big data which is internal and external data of the power grid in order to fully acquire data value and make business decisions more scientific.It is of great significance and value to improve the service level,reduce the operation cost and realize the value added of the information.This paper summarized the theory of the value added of the power consumption information.Firstly the paper introduced the concepts of value,information value and value added of information.Then the paper summarized the type and characteristics of the smart power consumption.Furthermore,the value level of the power user side information was decomposed and analyzed.Finally three kinds of information value added typical application scenes were proposed for different types of information.Based on the internal power data of the power grid and the typical application scene for abnormal electricity detection,a method based on user classification and Gaussian kernel density local outlier factor algorithm was proposed.Firstly,the users were classified by fuzzy clustering method.Then various features of each type of users were extracted and PCA was used to reduce the dimension of the feature vectors.Finally,we used the Gaussian kernel function to improve the local outlier factor(LOF)algorithm,and proposed the Gaussian kernel density local outlier factor(GKLOF)algorithm.The effectiveness of GKLOF algorithm was verified by the combination of theoretical analysis and simulation experiments.Based on the external data of the power grid and the typical application scene for power trading platform,this paper proposed an intelligent recommendation algorithm through mining the user's explicit and implicit preference information of the platform.And the information value added quantitative model was built to quantify the information value added.The experimental results showed that the improved algorithm still had high accuracy in the case of low volume of data at the early stage of the platform which can realize the information value added,and bring great benefits to the platform.