Rearch Of Smart Meter Data Analysis And Management Based On A Hierachical Model

The universality of smart meters brings a lot of electrical power data. A single data analysis model is noneffective to sovle the problems of the power industry because of those mass, high-frequecy, diversified power data. So the intergration of data analysis methods is needed for the power industry. How to analyze those electrical power data to gain practical value, has become a concern of the power industry.First of all, the thesis investigates the common methods and current applications of the data analysis in the power system. It also analyses the current problems faced by the power system. A hierarch model based on the tree structure and the AdaBoost regression model are established to solve those problems. Secondly, this thesis prensents the system architecture of smart meters management based on the hierarch model, which including SSH framework(Struts- Spring-Hibernate), requirements analysis, establishment and management of the hierarch model. In this part, a new method is raised to build the hierarch model without traversing a tree to get the whole children nodes of one node by connecting the preorder traversal number and the postorder traversal number. Based on this model, the management of smart meters is set up. And the update of the model when adding, deleting and modifying nodes is rechieved in the thesis. This model has solved the problems of locating stealing electricity point and fault area. Thirdly, the process of design and implementation of the smart meters management and data analysis system is presented. And the exhibition effect is showed on the website. Finanlly, this thesis builds the regression model of AdaBoost for short-term load forecasting. At the same time, the Support Vector Machine(SVM) algorithm is compared for experiments. The regression model of AdaBoost presents a better predition accuracy.The management of smart meters based on this hierarch model has solved monitoring stealing electricity and warning fault with data analysis. At the same time, the machine learning is used for shot-term load forecasting, providing data support for the power system, which has practical research significance in pwer industry.