摘要
考虑到无人机平台电力巡检有效载荷及计算能力有限,该文提出了一种基于无人机的智能电力巡检方案。首先,设计了无人机电力巡检软硬件框架;其次,提出了一种双阶段微调策略训练深度学习模型,从而提高模型应对特殊数据集的检测能力;最后,提出了一种通道剪枝方案,从而去除冗余的特征通道降低模型的复杂度,使得模型可有效部署于计算能力有限的无人机平台。实验阶段,以检测绝缘子为例,对所提方案进行验证。结果表明,所提双阶段微调策略可将基础网络性能提升约4%至10%。同时,所提通道剪枝方法可使得网络模型尺寸更小、性能更优。该模型为电力系统智能化及无人化监控管理提供了一定借鉴作用。
Considering the limited payload and computing power of the UAV platform,an intelligent power inspection scheme based on UAV is proposed.Firstly,the hardware and software framework of UAV power inspection is designed.Secondly,a two-stage fine-tuning strategy training depth learning model is proposed to improve the detection ability of the model for special data sets.Finally,a channel pruning scheme is proposed to remove redundant feature channels and reduce the complexity of the model,so that the model can be effectively deployed on the UAV platform with limited computing power.In the experimental stage,the proposed scheme is verified by taking the insulator detection as an example.The results show that the proposed two-stage fine-tuning strategy can improve the basic network performance by about 4%to 10%.At the same time,the proposed channel pruning method can make the network model smaller and have better performance.The model provides a reference for intelligent and unmanned monitoring and management of power system.
作者
鹿可可
李菱
欧发斌
万义飞
LU Ke-ke;LI Ling;OU Fa-bin;WAN Yi-fei(Guangxi Power Grid Co.,Ltd.,Nanning 530000,China)
出处
《自动化与仪表》
2023年第4期59-64,共6页
Automation & Instrumentation
关键词
智能电网
电力巡检
深度学习
通道剪枝
smart grid
electric power inspection
in-depth learning
channel pruning
