[1]何正军,石雪敏,杜 涛,等.基于深度残差网络的轻量化电力敏感数据处理模型设计[J].机械与电子,2026,44(04):114-118.
　HE Zhengjun,SHI Xuemin,DU Tao,et al.Design of Lightweight Processing Model for Power-sensitive Data Based on Deep Residual Network[J].Machinery & Electronics,2026,44(04):114-118.
点击复制

基于深度残差网络的轻量化电力敏感数据处理模型设计()
分享到：

《机械与电子》[ISSN:1001-2257/CN:52-1052/TH]

卷:
44
期数:
2026年04期
页码:
114-118
栏目:
电力控制
出版日期:
2026-04-27

文章信息/Info

Title:
Design of Lightweight Processing Model for Power-sensitive Data Based on Deep Residual Network
文章编号:
1001-2257 ( 2026 ) 04-0114-05
作者:
何正军; 石雪敏; 杜 涛; 王雪梅; 王佩霞
国网甘肃省电力公司天水供电公司,甘肃 天水 741000
Author(s):
HE Zhengjun ;  SHI Xuemin ;  DU Tao ;  WANG Xuemei ;  WANG Peixia
( Tianshui Power Supply Company , State Grid Gansu Electric Power Company , Tianshui 741000 , China )
关键词:
电力敏感数据; 深度残差网络; 负荷识别; 多尺度卷积
Keywords:
power-sensitive data ;  deep residual network ;  load identification ;  multi-scale convolution
分类号:
TM761
文献标志码:
A
摘要:
电力系统中的敏感数据具有多尺度波动、通道异构和时序耦合等特性,传统模型难以兼顾识别精度与结构轻量性。为此,提出一种轻量化深度残差网络 L-ResNet ,融合多尺度深度卷积、通道注意力门控与动态残差融合模块,构建统一的敏感数据处理框架。该模型通过多尺度建模与自适应融合提升特征表达能力,并在负荷异常识别、用户行为分类和敏感区间提取等任务中均表现出优于主流模型的性能。结果表明,L-ResNet 在保持高精度的同时显著降低了参数量与推理延迟,为电力敏感数据的高效处理与边缘部署提供了可行方案。
Abstract:
Sensitive data in electrical power systems typically exhibits characteristics such as multi scale fluctuations , channel heterogeneous , and strong temporal coupling , making it difficult for traditional models to balance recognition accuracy and structural lightweightness.To address this , a Lightweight Residual Network ( L-ResNet ) is proposed , integrating Multi-scale Depthwise Convolution ( MDC ), Channel Attention Gating ( CAG ), and Dynamic Residual Fusion ( DRF ) modules to construct a unified framework for processing sensitive data.This model enhances feature representation capability through multi scale modeling and adaptive fusion.It demonstrates superior performance over mainstream models across tasks including load anomaly identification , user behavior classification , and sensitive interval extraction. The results show that L-ResNet significantly reduces the number of parameters and inference latency while maintaining high accuracy , providing a feasible solution for the efficient processing and edge deployment of power-sensitive data.

参考文献/References:

[ 1 ] 林莉,范米,林雨露,等 . 基于不确定与扰动估计器的直流配电网电压鲁棒控制[ J ] . 电工技术学报,2023 , 38 ( 17 ): 4657-4671.

?[ 2 ] 石文娟,凌凡,张俊权,等 . 基于 ISSA 优化 EEMD 的用户侧电力负荷预测方法[ J ] . 信息技术, 2025 ( 2 ): 97-103.
[ 3 ] 江世雄,黄鸿标,陈苏芳,等 . 基于 GA 改进 LSTM BP神经网络的智慧楼宇用能行为预测方法[ J ] . 沈阳工业大学学报,2022 , 44 ( 4 ): 366-371.
[ 4 ] 王平 .Elman 神经网络在电力负荷预测中的应用探究[ J ] . 仪器仪表用户, 2024 , 31 ( 12 ): 108-110.
[ 5 ] 李凯鹏,宋晓阳,孙冉冉,等 . 电力终端侧网络异常入侵行为检测与预警技术研究[ J ] . 自动化仪表, 2025 , 46( 6 ): 11-15.
[ 6 ] 杨晓坤,杨晓波,陈雪敏,等 . 基于用电行为的电能计量异常实时检测[ J ] . 信息技术, 2025 ( 5 ): 185-189.
[ 7 ] 皮志勇,朱益,廖玄,等 . 基于深度学习的智能变电站通信链 路 故 障 定 位 方 法 [ J ] . 中 国 电力, 2023 , 56 ( 7 ):136-145.
[ 8 ] 孟磊 . 基于 CNN-LSTM 网络的电力负荷预测[ J ] . 机电信息,2023 ( 7 ): 7-10.
[ 9 ] XU J C , GAO L M , LIU Z H , et al.Research on fault diagnosis method of synchronous generator based on ResNet-18 [ C ] ∥Journal of Physics : Conference Series , 2025 , 2963 : 012021.
[ 10 ] QIN Y L , TANG Q Z , XIN J Z , et al.A rapid identification technique of moving loads based on MobileNetV2 and transfer learning [ J ] .Buildings , 2023 , 13( 2 ): 572.
[ 11 ] SALMAN D , FARAH A , ALI S A.Enhancing power grid stability through reactive power demand forecasting using deep learning [ J ] .International journal of electrical and electronics engineering , 2024 , 11 ( 12 ): 170-185.
[ 12 ] 吴亚丽,冯梦琦,王君虎,等 . 变分模态分解和自适应稀疏自编码器的故障诊断模型[ J ] . 机械科学与技术,2025 , 44 ( 9 ): 1603-1611.

备注/Memo

备注/Memo:
收稿日期: 2025-10-10
基金项目:国网甘肃省科学技术项目( B3270225Z359 )
作者简介:何正军 ( 1973- ),男,甘肃秦安人,硕士,高级工程师,研究方向为 电力系统数字 化架构设计与数 据安全应用研 究等;石雪敏 ( 1995- ),女,甘肃天水人,学士,工程师,研究方向为电力数据处理模型开发与优化等,通信作者, E-mail : WXHSJDJLDKS45@126.com 。
更新日期/Last Update: 2026-08-21