基于聚类的用户用电行为分析研究

发布时间：2018-05-02 00:05

本文选题：数据挖掘 + 需求响应　；参考：《华北电力大学(北京)》2017年硕士论文

【摘要】：随着用电信息采集系统的飞速发展,用电数据也日益增多,将数据挖掘技术应用于用电数据分析可以有效获取用户用电的相关规律与模式信息,支持用电服务个性化、差异化的需求,为用电智能化奠定基础。而如何有效地进行用电信息的数据挖掘以及使用数据挖掘所得信息支撑智能电网以及智能用电的建设成为迫切需要解决的问题。本文针对以上问题,从用户用电数据中隐藏着用户的用电行为习惯着手,使用聚类分析方法对这些用电数据进行挖掘并识别用户用电行为模式,并以此支撑用户智能用电需求响应策略。首先,本论文针对数据挖掘过程中的特征提取步骤进行优化,设计了一种基于信息熵与特征相关系数的用电特征优选策略,通过选出与用户用电模式关系紧密且独立性高的特征集来降低聚类计算过程的复杂性并改善聚类分析的效果。然后在特征优选的基础上使用一种初始聚类中心改善的k-mean聚类方法完成用电数据的聚类分析,并完成用户用电行为聚类分析的仿真实验,实验证明本文所提方法和策略能有效识别不同用电模式的用户。最后,本文使用聚类分析所得结果去实现智能用电需求响应策略的优化。通过针对不用用电模式的用户的分类调度,实现用户智能用电需求响应效果的优化,以及降低其计算的复杂度。同时方法基于分布式计算的思想,将优化过程拆解至各类用户,以充分利用用户资源实现互动调度过程,从而有效地提高计算效率并保护用户信息的安全性。
[Abstract]:With the rapid development of power information acquisition system and the increasing number of power data, the application of data mining technology in power data analysis can effectively obtain the relevant laws and mode information of users' electricity consumption, and support the individualization of electricity service. The demand of differentiation lays the foundation for the intelligent use of electricity. How to effectively mine the power information and how to use the data mining information to support the smart grid and the construction of smart electricity has become an urgent problem to be solved. In view of the above problems, this paper starts with the user's electric behavior habits hidden in the user's power consumption data, and uses the clustering analysis method to mine these data and to identify the user's electric behavior pattern. And to support the user intelligent demand response strategy. Firstly, this paper optimizes the feature extraction process in the process of data mining, and designs a strategy based on information entropy and feature correlation coefficient. In order to reduce the complexity of the clustering calculation process and improve the effect of clustering analysis, the feature sets which are closely related to the user power consumption mode and which are highly independent are selected to reduce the complexity of the clustering calculation process. Then, based on the feature selection, a k-mean clustering method with improved initial clustering center is used to complete the clustering analysis of power consumption data, and the simulation experiment of the user's electricity behavior clustering analysis is completed. Experimental results show that the proposed method and strategy can effectively identify users with different power consumption modes. Finally, the results of clustering analysis are used to optimize the intelligent demand response strategy. According to the classified scheduling of users without electricity consumption mode, the user intelligent demand response effect is optimized and the computational complexity is reduced. At the same time, based on the idea of distributed computing, the optimization process is broken down to all kinds of users, in order to make full use of user resources to realize interactive scheduling process, so as to improve the efficiency of computing and protect the security of user information effectively.
【学位授予单位】：华北电力大学(北京)
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13;TM73

【参考文献】