【作者】 彭谦；

【导师】 王振旗；

【作者基本信息】 华北电力大学 ， 工程硕士（专业学位）， 2018， 硕士

【摘要】 近些年国家大力推动实施电力需求侧管理,其中的一个重要环节就是有序用电管理。传统的有序用电方式都是硬性的调控,没有考虑用户的利益,如何充分考虑用户利益,科学化、精细化的制定有序用电政策,提高用户对有序用电管理的参与性、响应性,是目前长时间困扰着研究者的一个难题。本文基于国网公司现有的SG186和SG-ERP工程建设采集用户用电及社会用电数据,运用互联网和数据挖掘分析技术,设计开发了一个完整的基于Hadoop的有序用电管理系统,旨在维护用户利益,提升用电用户在参与有序用电时的体验和提高用户参与有序用电的积极性。系统分为负荷预测模块和有序用电管理模块。在负荷预测模块中,城市的用电负荷曲线用简明的折线图和地图模型显示日负荷特征和预测未来的短期电力负荷特征。该模块通过线性回归的方法建立短期电力负荷模型,对未来短时间内的电力负荷做预测。在有序用电管理模块,系统通过数据可视化技术实现了按地图形式查找数据,同时,该模块还能对用户的有序用电日报和有序用电方案进行详细查询。为了提高对用户用电特征进行聚类时的效率,本文结合Canopy聚类算法和K-means聚类算法,基于用户历史日负荷数据用Canopy算法求出聚类簇数,再用K-means算法聚类出四类典型的用户用电模式。总体架构上,系统采用Hadoop分布式架构,数据库采用Hive与Mysql相结合的方式,用以提高系统应对并发计算和运行的能力。更多还原

【Abstract】 In recent years,our country vigorously promotes the implementation of power demand-side management,the orderly electricity management is one of the important links.The traditional supervise way of orderly electricity is rigid,there are not consider users interests.How to fully take into account the interests of users,scientific and meticulous formulation of an orderly electricity policy,and improve user’s participation and responsiveness of orderly electricity management,it is a dilemma that has plagued researchers for so long.This paper collected user’s electricity data and social electricity data based on the SG186 and SG-ERP project construction of SGCC,designed and developed a complete Hadoop-based orderly electricity management system based on Internet and data mining analysis technology.This system aim at maintain the interests of users and improve the experience and enthusiam when users participating the orderly electricity.The system is divided into load forecasting module and orderly electricity management module.In the load forecasting module,the city’s electricity load curve shows the daily load characteristics and predicts the future short-term load characteristics with simple line charts and map models.This module use means of linear regression establish a short-term power load model to forecast power load which in a short period of time.In the orderly electricity management module,this system implement select data by map mode through the data visualization technology,at the same time,this module also can select the user’s orderly electricity daily report and orderly electricity program in detail.In order to improve the efficiency of clustering users’ electrical characteristics,this paper combines Canpoy clustering algorithm and K-means clustering algorithm,calculate the number of clusters based on user’s historical daily load data using Canopy algorithm,then use K-means algorithm clustering four types of typical user electricity patterns.On the overall system structure,system based on Hadoop distributed architecture,and the system database combina the Hive database and Mysql database to improve the system’s ability to cope with concurrent calculations and operations.更多还原