电力信息与通信技术2025，Vol.23Issue(7)：46-53,8.DOI:10.16543/j.2095-641x.electric.power.ict.2025.07.06
基于张量网络的电网视频数据差异化压缩技术
Differential Compression Technology for Power Grid Video Data Based on Tensor Networks
摘要
Abstract
The traditional unstructured data compression methods for power grid video use commercial and mature compression algorithms such as H.264/AVC,MPEG,H.265/HEVC,etc.These algorithms generally compress the video data indiscriminately.Although they can significantly reduce the space occupied by the video,they lead to the loss of details of key equipment,which seriously affects the accuracy of subsequent equipment state analysis.This paper innovatively utilizes the flexible calculation characteristics of tensor networks,introduces deep neural network machine learning algorithms,annotates and trains the main equipment of the power grid,such as transformers,etc.,and identifies the power equipment in the video frame in real time.Through tensor decomposition,the equipment and environment are quickly segmented,generating two new tensors and providing different compression treatments.This paper uses low compression rate Tucker tensor compression for power grid equipment data,and high compression rate Tucker tensor compression for non power grid equipment data,to achieve the function of reducing compression rate for important equipment while retaining precision,and increasing compression rate for environmental background to reduce storage space.So that video storage space is significantly reduced and details of major equipment are selectively retained.By balancing the business needs of reducing video footprint and precise business analysis,more flexible video data compression has been achieved.关键词
张量网络/神经网络/视频数据压缩/机器学习Key words
tensor networks/neural networks/video data compression/machine learning分类
信息技术与安全科学引用本文复制引用
徐敏,彭林,周爱华,朱力鹏,李尼格..基于张量网络的电网视频数据差异化压缩技术[J].电力信息与通信技术,2025,23(7):46-53,8.基金项目
国家电网有限公司总部科技项目资助"大规模电网高阶张量网络构建与计算方法研究"(5700-202358710A-3-3-JC). （5700-202358710A-3-3-JC）
