基于节点日负荷曲线的深度嵌入式聚类及其改进方法对比研究  

Comparative study on deep embedded clustering and its improved methods based on node daily load curve

作　　者：陈谦[1] 陈嘉雯 王苏颖 史锐 CHEN Qian;CHEN Jiawen;WANG Suying;SHI Rui(College of Energy and Elctrical Engineering,Hohai University,Nanjing 211100,China)

机构地区：[1]河海大学能源与电气学院,江苏南京211100

出　　处：《电力科学与技术学报》2023年第1期130-137,共8页Journal of Electric Power Science And Technology

基　　金：国家自然科学基金(51837004)。

摘　　要：基于日负荷曲线的负荷节点分类是负荷建模的重要环节,详略得当的分类结果保留了负荷节点的内在特性,可提升电力系统仿真计算的效率。当前基于人工智能的节点聚类方法进展迅速,然而总体上针对数据深层特征提取的适应性仍存在不足。采用了基于改进的深度嵌入式算法的日负荷曲线聚类方法,利用神经网络可有效提取数据的深层特征的能力。进而,提出一种先升维后聚类的改进方法,通过算例对比分析,验证了本文所提算法的可行性,以及所提升维—重构聚类方法的正确性。Load node classification based on daily load curve is an important part of load modeling.The detailed and appropriate classification results retain the internal characteristics of load nodes and can improve the efficiency of power system simulation calculation.At present,the node clustering method based on artificial intelligence has made rapid progress.However,the overall adaptability to data deep feature extraction is still insufficient.This paper presents the daily load curve clustering method based on the improved deep embedded algorithm,which uses the ability of neural network to effectively extract the deep features of the data.Then,an improved method of increasing the dimension first and then clustering is proposed.Through the comparative analysis of numerical examples,the feasibility of the proposed algorithm and the correctness of the improved dimension reconstruction clustering method are verified.

关 键 词：负荷建模 日负荷曲线聚类 深度嵌入式 升维-重构聚类 

分 类 号：TM712[电气工程—电力系统及自动化]