辽宁工程技术大学学报（自然科学版）2023，Vol.42Issue(6)：748-755,8.DOI:10.11956/j.issn.1008-0562.2023.06.015
基于改进Inception网络及LightGBM的弓网电弧识别方法
Pantograph-catenary arc recognition method based on improved Inception network and LightGBM
摘要
Abstract
Aiming at the problem of single data recognition in pantograph-catenary arc recognition method,a pantograph-catenary arc recognition method based on improved multi-scale convolutional network(GInception)combined with LightGBM is proposed.The original signal is denoised and reconstructed by using the combination of complete ensemble empirical mode decomposition(CEEMDAN)and wavelet soft threshold.The reconstructed signal is broadcasted and transformed into a multi-dimensional signal,which is input into the GInception network for feature extraction,and then the features extracted by the GInception network are imported into the lightweight gradient hoist for recognition.The results show that the recognition accuracy of pantograph-catenary arc reaches 96.3%under five different working conditions.The research conclusions can provide reference for the identification of pantograph-catenary arc in trams.关键词
弓网电弧/完整集合经验模态分解/卷积神经网络/多尺度卷积运算/轻量级梯度提升机Key words
pantograph catenary arc/CEEDMAN/convolutional neural network/multi-scale convolution operation/LightGBM分类
信息技术与安全科学引用本文复制引用
李斌,孙凤桐..基于改进Inception网络及LightGBM的弓网电弧识别方法[J].辽宁工程技术大学学报（自然科学版）,2023,42(6):748-755,8.基金项目
国家自然科学基金项目(51674136) （51674136）
