ADGWN: adaptive dual-channel graph wavelet neural network for topology identification of low-voltage distribution grid

Author:

Hu Kekun1,Zhu Zheng2,Xu Yukun2,Jiang Chao2,Dai Chen3

Affiliation:

1. Inspur Electronic Information Industry Co., Ltd., Jinan, China

2. Power Research Institute, State Grid Shanghai Municipal Electric Power Company, Shanghai, China

3. State Grid Shanghai Municipal Electric Power Company, Shanghai, China

Abstract

Maintaining accurate topology of the low-voltage distribution grid (LVDG) are critical to the operations and maintenance of power distribution systems. However, this goal is hard to achieve due to the fast-changing LVDG topology. To this end, we focus on the abnormal customer-transformer relationships identification in the LVDG and propose an identification method based on an Adaptive Dual-channel Graph Wavelet Neural Network (ADGWN) consisting of two identical GWNs connected with the attention mechanism. In the proposed ADGWN, two GWNs learn customer embedding simultaneously from the LVDG topology graph and the feature graph that is constructed from customer electricity consumption data with the k-Nearest Neighbor algorithm. The topology identification results of these two GNNs are then adaptively fused to form the ultimate identification result with the attention mechanism by dynamically balancing the aforementioned two types of information. To validate the performance of our proposed method, we further build a real benchmarking dataset from customer electricity consumption data collected from a certain substation in Shanghai, China. Experimental results show that the proposed ADGWN achieves 100.0% LVDG topology identification accuracy and significantly outperforms the state-of-the-art. Our proposed method can help operators of power distribution systems maintain the accurate topology in a timely and economic manner.

Publisher

IOS Press

Subject

Artificial Intelligence,General Engineering,Statistics and Probability

Reference21 articles.

1. Identifying topology of low voltage distribution networks based on smart meter data;Pappu;IEEE Transactions on Smart Grid,2017

2. Identification of distribution network topology parameters based on multidimensional operation data;Li;Energy Reports,2021

3. Topology identification of distribution networks using a split-EM based data-driven approach;Ma;IEEE Transactions on Power Systems,2021

4. Analysis and applications of smart meter data;Xu;Utilization & Distribution,2015

5. A learning-to-infer method for real-time power grid multi-line outage identification;Zhao;IEEE Transactions on Smart Grid,2019

Cited by 1 articles. 订阅此论文施引文献 订阅此论文施引文献,注册后可以免费订阅5篇论文的施引文献,订阅后可以查看论文全部施引文献

同舟云学术

1.学者识别学者识别

2.学术分析学术分析

3.人才评估人才评估

"同舟云学术"是以全球学者为主线,采集、加工和组织学术论文而形成的新型学术文献查询和分析系统,可以对全球学者进行文献检索和人才价值评估。用户可以通过关注某些学科领域的顶尖人物而持续追踪该领域的学科进展和研究前沿。经过近期的数据扩容,当前同舟云学术共收录了国内外主流学术期刊6万余种,收集的期刊论文及会议论文总量共计约1.5亿篇,并以每天添加12000余篇中外论文的速度递增。我们也可以为用户提供个性化、定制化的学者数据。欢迎来电咨询!咨询电话:010-8811{复制后删除}0370

www.globalauthorid.com

TOP

Copyright © 2019-2024 北京同舟云网络信息技术有限公司
京公网安备11010802033243号  京ICP备18003416号-3