基于合成时频分布图的轻量级网络小电流故障电弧检测  

Detection of Small Current Fault Arc in Lightweight Network Based on Synthetic Time-frequency Distribution Graphs

作　　者：冷友伟 LENG Youwei(Songjiang Fire and Rescue Division,Shanghai 201600,China)

机构地区：[1]上海市松江区消防救援支队,上海201600

出　　处：《中国人民警察大学学报》2024年第8期54-60,共7页Journal of China People's Police University

摘　　要：为有效解决故障电弧检测问题,提出一种基于时频分布图的轻量级网络小电流故障电弧检测方法。参照相关标准进行电弧试验,并采集电弧试验数据,通过把电流数据转换成合成时频分布图构造训练集和测试集,输入STF-GhostNet模型识别故障电弧并输出结果。试验结果表明:采用该方法进行故障电弧检测准确率约为94.1%,与传统BP模型、AlexNet相比,准确率明显提高。In order to effectively solve the problem of arc fault detection,a method of low-current fault arc detection in lightweight networks based on time-frequency distribution graphs is proposed.An arc experiment was conducted based on related standard to collect arc experiment data,thereby constructing a training set and test set by convert⁃ing current data into synthetic time-frequency distribution graphs.STF-GhostNet model was used to identify arc fault and output the result.Experimental results show that the accuracy of arc fault detection using this method is about 94.1%,which is higher than the traditional BP model and AlexNet.

关 键 词：神经网络 时频分析 电弧故障检测 小电流电弧 STF-GhostNet 

分 类 号：TU998.1[建筑科学—市政工程]