Research And Implementation Of Power Big Data Visualization Based On Hadoop

With the continuous development of smart grid construction,the scale of modern power systems is becoming increasingly large.The power data acquired by each acquisition system has the characteristics of large amount of data,diverse types,and low degree of resource integration.How to analyze and realize data visualization of massive power data is of great significance for the safe and stable operation of power systems.With the mature application of computer technology in power system,the power data with large amount of information can be accommodated by the graphics and image resources,which provides a theoretical support for the visualization of large data.Therefore,this thesis takes the power big data as the research object,conducts abnormal electricity behavior detection and power load prediction for power big data based on Hadoop platform,and realizes the visualization of power big data.This thesis mainly completed the following work.1.Combining the massive multi-source characteristics of power data,the demand analysis of power big data visualization was conducted.The overall architecture of power big data visualization based on Hadoop was designed,and the key technologies involved were briefly described.2.A method to detect abnormal electricity behavior of users is proposed.Through the extraction of feature schemes from power consumption data,and combined with the characteristics of electricity consumption characteristics of major criminal activities,the abnormal electricity behavior is analyzed,and the abnormal power users are determined according to the early warning condition.MapReduce in the Hadoop platform is used for parallel computing,and the calculation results are combined with geographic information to realize the visualization of abnormal electrical behavior based on geographical information.3.A wavelet neural network power load forecasting method based on FCM is proposed.The power users are personally classified by FCM clustering method.The load forecasting of the classified data is carried out by wavelet neural network and parallelized on distributed platform.Finally,the parallelized calculation results of load forecasting are visualized.Compared with the commonly used load data time series prediction method,this method has significantly improved the prediction accuracy.Finally,the Hadoop large data experiment platform is built,and the experimental results of the different common electrical behavior detection and the power load forecasting method are visualized by Echarts,and the performance of the large data cluster is tested and analyzed.The experimental results show that the study of Hadoop-based power big data visualization can meet the actual application requirements.