大模型时代:电力视觉技术新起点

扫码查看

原文链接

国家科技期刊平台
NETL
NSTL
万方数据

中文摘要：随着无人机、巡检机器人和远程监控系统在输电、变电、配电、安监等电力场景中的广泛应用,利用电力视觉技术完成对海量巡检图像的自动处理,能够进一步提升电力系统智能化运维水平,对我国源网荷储一体化进程的快速推进具有至关重要的作用.随着通用视觉大模型的兴起,电力视觉技术正处于从传统深度学习时代向大模型时代跨越的重要节点.该文首先综述了电力视觉技术和通用视觉大模型的最新研究进展,结合视觉大模型在多种公共场景的应用先例,探讨视觉大模型在电力视觉领域将面临的 3 重能力边界问题.从初步探索通用视觉大模型的潜力,到逐步构建电力视觉大模型的过程,提出 4 种模型应用范式以突破视觉大模型能力边界.最后分析了视觉大模型对电力视觉研究者的影响,并对大模型浪潮下电力视觉技术的发展方向进行了展望.

外文标题：The Era of Large Models:A New Starting Point for Electric Power Vision Technology

外文摘要：With the widespread applications of drones,inspection robots,and remote monitoring systems in power sce-narios such as transmission,transformation,distribution,and safety supervision,the power vision technology is utilized to automatically process massive inspection images,thus the intelligent operation and maintenance level of power systems can be further enhanced,playing a crucial role in the rapid advancement of China's source-grid-load-storage integration process.With the rise of general-purpose large vision models,the electric power vision technology is embarking on an important transition from the traditional deep learning era to the era of large models.In this paper,the latest research pro-gress in electric power vision technology and general-purpose large vision models is summarized.Combined the precedents of large vision models'applications in various public scenarios,three main capability boundary issues that large vision models will face in the field of power vision are investigated.From initially attempting to apply gen-eral-purpose large vision models to establishing power-specific large vision models,the four model application paradigms to break through the capability boundaries of large vision models are proposed.Finally,the impact of large vision models on electric power vision researchers is analyzed,and prospects in the development direction of electric power vision technology under the wave of large models are put forward.

外文关键词：

electric power visionlarge vision modelsobject detectionimage segmentationdeep learningimage pro-cessing

作者：

赵振兵、冯烁、席悦、张靖梁、翟永杰、赵文清

展开 >

作者单位：

华北电力大学电气与电子工程学院,保定 071003

华北电力大学复杂能源系统智能计算教育部工程研究中心,保定 071003

华北电力大学河北省电力物联网技术重点实验室,保定 071003

华北电力大学控制与计算机工程学院,保定 071003

展开 >

关键词：

电力视觉视觉大模型目标检测图像分割深度学习图像处理

基金：

国家自然科学基金国家自然科学基金国家自然科学基金国家自然科学基金河北省自然科学基金河北省自然科学基金中央高校基本科研业务费专项

项目编号：

U21A20486623731516237111862303184F2021502008F20215020132023JC006

出版年：

2024

DOI：

10.13336/j.1003-6520.hve.20232149

高电压技术

中国电力科学研究院　中国电机工程学会

高电压技术

CSTPCD北大核心

影响因子：2.32

ISSN：1003-6520

年,卷(期)：2024.50(5)