Classification of electric vehicle charging time series with selective clustering
Classification of electric vehicle charging time series with selective clustering
复制标题
DOI：
10.1016/j.epsr.2020.106695
复制
发表时间：
2020-12
影响因子：
3.9
通讯作者：
Chenxi Sun;Tongxin Li;S. Low;V. Li
中科院分区：
文献类型：
--
作者：
Chenxi Sun;Tongxin Li;S. Low;V. Li
我们开发了一种新的迭代聚类方法分类电动汽车充电率的时间序列的基础上，他们的“尾部功能”。我们的方法首先从具有不同长度、包含缺失数据并且被调度算法和测量噪声扭曲的多样性的充电时间序列中提取尾部。然后将收费尾部聚类为少量类型，然后使用其代表来改进尾部提取。这个过程迭代直到收敛。我们将我们的方法应用于最近公开的细粒度电动汽车充电数据集ACN-Data，以说明其有效性和潜在应用。
We develop a novel iterative clustering method for classifying time series of EV charging rates based on their “tail features”. Our method first extracts tails from a diversity of charging time series that have different lengths, contain missing data, and are distorted by scheduling algorithms and measurement noise. The charging tails are then clustered into a small number of types whose representatives are then used to improve tail extraction. This process iterates until it converges. We apply our method to ACN-Data, a fine-grained EV charging dataset recently made publicly available, to illustrate its effectiveness and potential applications.