摘要
在低压配电系统中,准确的户变关系识别对系统运行和管理至关重要。文章提出了一种基于深度高斯混合模型的自动户变识别算法,该算法基于噪声标签与半监督学习理论,充分利用已有标签信息,通过聚类主网络以及分割聚合子网络结构的设计,实现了台区数量未知场景下的建档或校核。在基于实际用户数据的户变关系识别实验中,所提算法在无标记用户建档的准确率高达98.8%,与传统聚类和半监督深度学习算法相比,在各类工况下都取得了出色的性能。
In low-voltage distribution systems,accurate identification of customer-transformer relationships is crucial for system operation and management.This paper proposes an automatic customer-transformer relationship identification algorithm based on a deep Gaussian mixture model,which leverages noise-labels and semi-supervised learning theory.By designing a clustering main network and a splitting and merging subnetwork structure,it fully utilizes existing label in-formation to achieve document establishment or verification in scenarios where the number of substations is unknown.In experiments on identifying customer-transformer'relationships based on actual user data,the proposed algorithm achieves an accuracy of up to 98.8%in profiling unlabeled users,demonstrating superior identification performance compared to traditional clustering and semi-supervised deep learning algorithms under various operating conditions.
作者
郭祥葛
吕志宁
李毅
郑杰
史万福
王辉
GUO Xiangge;LYU Zhining;LI Yi;ZHENG Jie;SHI Wanfu;WANG Hui(Shenzhen Power Supply Bureau Co.,Ltd.,Shenzhen 518000,Guangdong Province,China;State Grid Electric Power Research Institute Co.,Ltd.,Nanjing 210061,Jiangsu Province,China)
出处
《电力信息与通信技术》
2025年第11期51-58,共8页
Electric Power Information and Communication Technology
基金
中国南方电网公司科技项目“基于负荷特征传导的低压全景成图技术研究及应用”(090000KK52222164)。
关键词
混合高斯
户变关系识别
自适应聚类
低压配电系统
GMM
customer-transformer relationship identification
adaptive clustering
low-voltage distribution system
