摘要
提出一种基于随机森林分类的直流微电网孤岛检测方法。该方法首先对原始数据进行清洗并提取特征,选择直流母线侧的电压、电流、输出有功功率及3者的一阶后向差分等6个孤岛特性指标作为检测特征,生成特征向量集,然后基于随机森林分类建立直流微电网的孤岛检测模型,实现了孤岛的准确检测,最后与决策树分类法进行比较,随机森林分类法在处理大量数据情况下可更加准确地检测孤岛。
When the distributed generation system is connected to the grid,it is in an islanding state,which affects the safe and regular operation of the power system.The anti-islanding device must detect the islanding within an acceptable time limit.This paper proposes a DC microgrid islanding detection method based on random forest classification,first,to clean the original data and extract the features.Six islanding characteristics:voltage,current,output active power,and the first-order backward difference of the three are selected as the detection features to generate the feature vector set.Then,the islanding detection model of DC microgrid is built based on the random forest classification to achieve the accurate detection of islanding.Finally,compared with decision tree classification,random forest classification can detect islanding more accurately when it is processing large amounts of data.
作者
万庆祝
吴开聪
Wan Qingzhu;Wu Kaicong(School of Electrical and Control Engineering,North China University of Technology,Beijing 100144,China)
出处
《太阳能学报》
EI
CAS
CSCD
北大核心
2022年第1期269-276,共8页
Acta Energiae Solaris Sinica
基金
北京市电力节能协同创新(2011)研究项目(PXM2018_014212_000015_4_4)。
关键词
孤岛检测
随机森林
直流微电网
数据清洗
特征提取
islanding detection
random forest
DC micro-grid
feature extraction
data cleaning