Protection Algorithm Based on Two-Dimensional Spatial Current Trajectory Image and Deep Learning for Transmission Lines Connecting Photovoltaic Stations

Fiber differential protection (FDP) is the primary protection scheme in power systems. However, with the increasing proportion of photovoltaic (PV) grids connected in the power system, the controllability and weak power supply characteristics of photovoltaic power stations change the amplitude and p...

Full description

Saved in:
Bibliographic Details
Main Authors: Panrun Jin, Jianling Liao, Wenqin Song, Xushan Zhao, Yankui Zhang
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Applied Sciences
Subjects:
Online Access:https://www.mdpi.com/2076-3417/15/11/6066
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850129369729597440
author Panrun Jin
Jianling Liao
Wenqin Song
Xushan Zhao
Yankui Zhang
author_facet Panrun Jin
Jianling Liao
Wenqin Song
Xushan Zhao
Yankui Zhang
author_sort Panrun Jin
collection DOAJ
description Fiber differential protection (FDP) is the primary protection scheme in power systems. However, with the increasing proportion of photovoltaic (PV) grids connected in the power system, the controllability and weak power supply characteristics of photovoltaic power stations change the amplitude and phase angle characteristics of fault currents, which makes the sensitivity of fiber differential protection decline and even increases the risk of failure to operate. In view of this phenomenon, combined with the digital and intelligent development of the new energy power system, this study integrates deep learning with relay protection to propose a protection algorithm based on a two-dimensional spatial current trajectory image and deep learning. In this algorithm, the PV side current and the system side current are, respectively, mapped to the two-dimensional space plane as X- and Y-axes to form the current trajectory image. Under different fault conditions, they have obvious differences. A data-enhanced convolutional neural network (A-CNN) based on cross-overlapping data sets is used to identify trajectory features and locate faults. After performance evaluation, the protection algorithm has the advantages of adapting to new energy access, resisting transition resistance, and robustness to current transformer (CT) saturation, and outliers.
format Article
id doaj-art-0fdbcd16efa54015a88e5a481f978a55
institution OA Journals
issn 2076-3417
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Applied Sciences
spelling doaj-art-0fdbcd16efa54015a88e5a481f978a552025-08-20T02:33:01ZengMDPI AGApplied Sciences2076-34172025-05-011511606610.3390/app15116066Protection Algorithm Based on Two-Dimensional Spatial Current Trajectory Image and Deep Learning for Transmission Lines Connecting Photovoltaic StationsPanrun Jin0Jianling Liao1Wenqin Song2Xushan Zhao3Yankui Zhang4Economic and Technical Research Institute of Gansu Electric Power Company, Lanzhou 730050, ChinaSchool of Mechanical and Electrical Engineering, China University of Mining & Technology-Beijing, Beijing 100083, ChinaEconomic and Technical Research Institute of Gansu Electric Power Company, Lanzhou 730050, ChinaEconomic and Technical Research Institute of Gansu Electric Power Company, Lanzhou 730050, ChinaSchool of Mechanical and Electrical Engineering, China University of Mining & Technology-Beijing, Beijing 100083, ChinaFiber differential protection (FDP) is the primary protection scheme in power systems. However, with the increasing proportion of photovoltaic (PV) grids connected in the power system, the controllability and weak power supply characteristics of photovoltaic power stations change the amplitude and phase angle characteristics of fault currents, which makes the sensitivity of fiber differential protection decline and even increases the risk of failure to operate. In view of this phenomenon, combined with the digital and intelligent development of the new energy power system, this study integrates deep learning with relay protection to propose a protection algorithm based on a two-dimensional spatial current trajectory image and deep learning. In this algorithm, the PV side current and the system side current are, respectively, mapped to the two-dimensional space plane as X- and Y-axes to form the current trajectory image. Under different fault conditions, they have obvious differences. A data-enhanced convolutional neural network (A-CNN) based on cross-overlapping data sets is used to identify trajectory features and locate faults. After performance evaluation, the protection algorithm has the advantages of adapting to new energy access, resisting transition resistance, and robustness to current transformer (CT) saturation, and outliers.https://www.mdpi.com/2076-3417/15/11/6066convolutional neural networkcurrent trajectoryCT saturationfault location
spellingShingle Panrun Jin
Jianling Liao
Wenqin Song
Xushan Zhao
Yankui Zhang
Protection Algorithm Based on Two-Dimensional Spatial Current Trajectory Image and Deep Learning for Transmission Lines Connecting Photovoltaic Stations
Applied Sciences
convolutional neural network
current trajectory
CT saturation
fault location
title Protection Algorithm Based on Two-Dimensional Spatial Current Trajectory Image and Deep Learning for Transmission Lines Connecting Photovoltaic Stations
title_full Protection Algorithm Based on Two-Dimensional Spatial Current Trajectory Image and Deep Learning for Transmission Lines Connecting Photovoltaic Stations
title_fullStr Protection Algorithm Based on Two-Dimensional Spatial Current Trajectory Image and Deep Learning for Transmission Lines Connecting Photovoltaic Stations
title_full_unstemmed Protection Algorithm Based on Two-Dimensional Spatial Current Trajectory Image and Deep Learning for Transmission Lines Connecting Photovoltaic Stations
title_short Protection Algorithm Based on Two-Dimensional Spatial Current Trajectory Image and Deep Learning for Transmission Lines Connecting Photovoltaic Stations
title_sort protection algorithm based on two dimensional spatial current trajectory image and deep learning for transmission lines connecting photovoltaic stations
topic convolutional neural network
current trajectory
CT saturation
fault location
url https://www.mdpi.com/2076-3417/15/11/6066
work_keys_str_mv AT panrunjin protectionalgorithmbasedontwodimensionalspatialcurrenttrajectoryimageanddeeplearningfortransmissionlinesconnectingphotovoltaicstations
AT jianlingliao protectionalgorithmbasedontwodimensionalspatialcurrenttrajectoryimageanddeeplearningfortransmissionlinesconnectingphotovoltaicstations
AT wenqinsong protectionalgorithmbasedontwodimensionalspatialcurrenttrajectoryimageanddeeplearningfortransmissionlinesconnectingphotovoltaicstations
AT xushanzhao protectionalgorithmbasedontwodimensionalspatialcurrenttrajectoryimageanddeeplearningfortransmissionlinesconnectingphotovoltaicstations
AT yankuizhang protectionalgorithmbasedontwodimensionalspatialcurrenttrajectoryimageanddeeplearningfortransmissionlinesconnectingphotovoltaicstations