On the distinctiveness of the electricity load profile
On the distinctiveness of the electricity load profile
复制标题
DOI：
10.1016/j.patcog.2017.09.039
复制
发表时间：
2018-02
期刊：
影响因子：
--
通讯作者：
M. Bicego;A. Farinelli;E. Grosso;Dimitri Paolini;S. Ramchurn
中科院分区：
文献类型：
--
作者：
M. Bicego;A. Farinelli;E. Grosso;Dimitri Paolini;S. Ramchurn
最近越来越多的细粒度电力消耗数据的可用性允许利用模式识别技术来表征和分析能源客户的行为。模式识别分析通常在组级别执行，即目的是通过聚类技术发现具有一致行为的用户组-这对于例如有针对性的定价或集体能源购买是有用的。在本文中，我们沿着这一方向向前沿着了一步，研究了区分单个用户行为的可能性-即，从生物识别学的角度来看这方面的问题尚未得到适当解决，将为关键业务铺平道路，例如根据行为目标制定替代广告计划。为了研究负荷曲线（即每日电能消耗）的独特性，在我们的研究中，我们使用了原始数据（原始能耗时间序列）以及不同类型的特征，如频率系数和归一化负荷形状指数，以及各种分类方案。在两个真实的世界数据集上获得的结果表明，负载配置文件确实包含关于单个用户的显著的独特信息。
The recent increasing availability of fine-grained electrical consumption data allows the exploitation of Pattern Recognition techniques to characterize and analyse the behaviour of energy customers. The Pattern Recognition analysis is typically performed at group level, i.e. with the aim of discovering, via clustering techniques,groups of userswith a coherent behaviour – this being useful, for example, for targeted pricing or collective energy purchasing. In this paper we took a step forward along this direction, investigating the possibility of discriminating the behaviours ofsingle users– i.e., in a biometrics sense. This aspect has not been properly addressed and would pave the way to crucial operations, such as the derivation of alternative advertising schemes based on behavioural targeting. To investigate the uniqueness of the load profiles (i.e. the daily consumption of electrical energy), in our study we used the raw data (the original energy consumption time series) as well as different types of features such as frequency coefficients and normalized load shape indexes, together with various classification schemes. Results obtained on two real world datasets suggest that the load profile does contain significant distinctive information about the single user.