A topology identification method of low-voltage distribution network based on structured manifold and density peaks clustering

The consumer-transformer and consumer-phase relationships in low voltage distribution networks (LVDN) are often wrong and change frequently. A topology identification method of LVDN based on the structured manifold and density peak clustering (SM-DPC) is proposed. The method contains two stages, fea...

Full description

Saved in:
Bibliographic Details
Published in:2022 IEEE 2nd International Conference on Data Science and Computer Application (ICDSCA) pp. 59 - 63
Main Authors: Yan, Shaokui, Ding, Haili, Hou, Rui
Format: Conference Proceeding
Language:English
Published: IEEE 28.10.2022
Subjects:
Clustering algorithms
density peak clustering
distribution grid
Low voltage
manifold learning
Manifolds
Network topology
Redundancy
Shape
topology identification
Voltage
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The consumer-transformer and consumer-phase relationships in low voltage distribution networks (LVDN) are often wrong and change frequently. A topology identification method of LVDN based on the structured manifold and density peak clustering (SM-DPC) is proposed. The method contains two stages, feature selection and feature clustering. In the feature selection stage, the structured manifold learning algorithm is used to compute the low-dimensional graph embedding of voltage data, which reduces feature redundancy and maintains the original arbitrary distribution features. In the feature clustering stage, based on the feature selection results, the topology identification is achieved by clustering similar low-dimensional manifolds in the same cluster and different low-dimensional manifolds in other clusters through the density peak clustering algorithm (DPC). The example results show that the SM-DPC is more accurate in discovering clustering centers than the original DPC algorithm, and can improve the accuracy of clustering where the data distribution presents arbitrary shapes. Therefore, the proposed topology identification method has higher accuracy for recognizing consumer-transformer and consumer-phase relationships in LVDN.
DOI:10.1109/ICDSCA56264.2022.9988265