Research And Implementation Of Mining System For User Power Consumption Behavior

Due to the reform of power system,the old business model of extensive management and unified customer service is hard to deal with the growing demand of personalized and accurate customer service experience.So the most important things for the electricity company are how to make use of data resources to dig deeper into customer's potential needs and improve the quality of power services.After years of development and precipitation,the current national grid has accumulated massive service information,power grid production and other data,which can effectively support the massive power data analysis.Therefore,the State Grid Corporation hopes to carry out customer analysis through big data analysis technology to support targeted and meticulous customer service strategy.Under the urgent need of grid big data,the 863 grid big data project officially started and our system is designed and developed under the sponsor of the 863 Power Grid Big Data Project.The main goal is to design and implement the user's data analysis system of electrical power data.The system has some functions such as user's behavior classification,behavior evolutionary clustering analysis and abnormal behavior analysis along with big data methods.Compared with other grid data processing platforms,the system has the ability to handle big data and has a friendly user interface.What's more,the system has the ability to analyze power consumption trend and abnormal behavior which are new and meaningful.To verify the feasibility of the system,this article research related technologies by checking out a lot of literature.Then demand analyze and key technical analyze are made.As for key technical analyze,two algorithms are promote.First,in order to make the data model better,plenty of data preprocessing is made and this article promotes an algorithm named'BIC Model Selection Method Based On Backward Selection' which can reduce model complexity and improve model complexity.The other algorithm is used to improve power consumption analyze.The original algorithm used for power consumption analyze is evolutionary clustering which has some disadvantages that will lead to the decline in clustering quality.Plenty of experiments and some journals have proved all these two algorithms useful.With the solution of the key problem,this article carried out the overall system design and detailed design and finally build the fully functional system with a user friendly UI.Then tests are made to insure the reliability of the system.Finally,this article makes a summary and raises the shortcomings to be solved in the future.