摘要
时间序列处理在新能源出力分析、电力设施容量规划、系统风险评估等方面具有重要作用。然而,由于缺乏合理的数据划分方案,难以充分衡量多维数据间的关联。鉴于此,提出了一种基于shapelet变换的新能源电站出力多维数据可解释划分算法,直观给出场景划分依据。首先,基于原始时序信息以及其一阶差分,通过符号化近似算法进行多维时间序列相似性统计,进行shapelet预筛选;其次,建立了基于距离的shapelet评价标准,选取最优shapelet建立决策树,进而达到多维场景划分的目的;最后,采用某实际新能源电站数据开展算例分析。结果表明,所提方法可大幅降低多维数据聚类的计算复杂度,增强新能源电站多维数据划分的可解释性。
Time series processing plays an important role in new energy output analysis,power facility capacity planning,and system risk assessment.However,due to the lack of a reasonable and interpretable data partitioning scheme,it is difficult to fully measure the correlation between multidimensional data.In view of this,a multidimensional data interpretable partitioning algorithm for new energy power plant output based on shapelet transformation was proposed,providing intuitive basis for scene partitioning.Firstly,based on the original time series information and its first-order differences,a Symbolic ApproXimation(SAX)algorithm was used to perform multidimensional time series similarity statistics and perform shapelet pre screening.Secondly,a distance-based shapelet evaluation criterion was established to select optimal shapelet for constructing a decision tree,thereby achieving the goal of multidimensional scenario partitioning.Case studies were conducted using actual new energy power station data.Results show that the proposed method significantly can reduce the computational complexity of multidimensional data clustering and enhance the interpretability of multidimensional data partitioning for the new energy power station.
作者
胡桂荣
唐华
程士东
李檀
张强
Hu Guirong;Tang Hua;Cheng Shidong;Li Tan;Zhang Qiang(State Grid Yancheng Power Supply Company,Yancheng 224005,China)
出处
《能源与环保》
2025年第6期203-210,共8页
CHINA ENERGY AND ENVIRONMENTAL PROTECTION
基金
国网江苏省电力有限公司科技项目(J2023165)。
关键词
多维时间序列
符号化变换
shapelet
新能源电站
可解释数据划分
multidimensional time series
symbolic transformation
shapelet
new energy power station
interpretable data partitioning
