基于光伏系统发现未知工况和物理含义的聚类算法  

Clustering algorithm based on PV system for discovering unknown working conditions and physical meaning

作　　者：李小坤 刘光宇[1] 俞玮捷 俞武嘉[1] LI Xiaokun;LIU Guangyu;YU Weijie;YU Wujia(School of Automation Engineering,Hangzhou Dianzi University,Hangzhou 310000,China)

机构地区：[1]杭州电子科技大学自动化学院,浙江杭州310000

出　　处：《传感器与微系统》2021年第10期143-146,共4页Transducer and Microsystem Technologies

基　　金：国家自然科学基金资助项目(61174074);国家重大科研仪器研制资助项目(61427808)。

摘　　要：在光伏领域,光强以外的高维时间序列数据如何聚类,时间序列聚类是否具有明确的物理意义,目前还没有研究。时间序列聚类是一种从复杂、海量的时间序列数据集中提取有价值信息的新技术。光伏电站在不同工况下运行,统计电流、电压、温度和辐照度等时间序列数据,然后通过时序DBSCAN算法对时间序列数据聚类。实验结果表明:未知工况与聚类结果之间确实存在物理关联。未知工况与时间序列簇之间物理意义的发现,将推动光伏发电系统监测、预测和故障诊断的智能技术向前发展。In the field of photovoltaic,how to cluster high-dimensional time series data other than light intensity and whether time series clustering has clear physical significance have not been studied.Time series clustering is a new technique to extract valuable information from complex and massive time series data sets.The photovoltaic power station operates under different working conditions,and time series data such as current,voltage,temperature and irradiance are statistics.Then,the time series data are clustered through the timing DBSCAN algorithm.The experimental results show that there is a physical correlation between the unknown conditions and the clustering results.The discovery of the physical meaning between the unknown working conditions and time series clusters will promote the development of intelligent technologies for monitoring,forecasting and fault diagnosis of photovoltaic power generation systems.

关 键 词：光伏发电系统 时序聚类 DBSCAN算法 动态时间规整 

分 类 号：TP181[自动化与计算机技术—控制理论与控制工程]