基于MWOA-Elman神经网络的接地网瞬变电磁缺陷识别  

Transient Electromagnetic Defect Identification of Grounding Grid Based on MWOA-Elman Neural Network

作　　者：韩欣月 邓长征[1,2] 付添 夏鹏雨 刘旋 HAN Xinyue;DENG Changzheng;FU Tian;XIA Pengyu;LIU Xuan(College of Electrical and New Energy,China Three Gorges University,Yichang Hubei 443002,China;Hubei Provincial Collaboration Innovation Center for New Energy Microgrid,China Three Gorges University,Yichang Hubei 443002,China)

机构地区：[1]三峡大学电气与新能源学院,湖北宜昌443002 [2]三峡大学新能源微电网湖北省协同创新中心,湖北宜昌443002

出　　处：《广西师范大学学报（自然科学版）》2023年第3期53-66,共14页Journal of Guangxi Normal University:Natural Science Edition

基　　金：国家自然科学基金(51877122);湖北省教育厅科学技术研究计划中青年人才项目(Q20151205)。

摘　　要：为改进当前瞬变电磁探测系统的局限,提高接地网缺陷识别的效率与精度,提出一种MWOA-Elman神经网络,完成数据由采样到成像的转化过程,快速实现视电阻率成像,精准识别接地网的不同缺陷。首先,通过理论计算完成接地网瞬变场参数样本集,构造Elman神经网络的单映射关系。其次,围绕收敛因子、自适应权重与阈值对鲸鱼算法进行改进,用改进鲸鱼算法(modified whale optimization algorithm, MWOA)优化Elman神经网络的权值和阈值。测试结果表明,MWOA-Elman神经网络在第854步收敛,4项误差指标MAE、MSE、RMSE、MAPE分别为0.103 51、0.040 09、0.126 64和0.333 52%,接地网缺陷识别精度为99.678%,识别效率与精度均优于其他模型。最后,通过分析3×3接地网3种典型缺陷位置的成像结果,验证了MWOA-Elman神经网络应用于接地网缺陷识别的有效性,为嵌入瞬变电磁探测系统的智能算法提供参考。In order to improve the limitations of the current transient electromagnetic detection system and improve the efficiency and accuracy of ground grid defect identification,a MWOA-Elman neural network is proposed to complete the transformation process from sampling to imaging,quickly realize the apparent resistivity imaging,and accurately identify different defects of the ground grid.Firstly,the transient field parameter sample set of grounding grid is completed through theoretical calculation,and the single mapping relation of Elman neural network is constructed.Secondly,the modified Whale algorithm is improved around convergence factor,adaptive weight and threshold,and the weight and threshold of Elman neural network are optimized by modified whale optimization algorithm(MWOA).The test results show that the MWOA-Elman neural network converges at the 854 step,and the four error indexes MAE,MSE,RMSE,MAPE are 0.10351,0.04009,0.12664 and 0.33352%,respectively.The identification accuracy of grounding grid defects is 99.678%,and the identification efficiency and accuracy are better than that of other models.Finally,the imaging results of three typical defect locations of 3×3 grounding grid can verify the effectiveness of MWOA-Elman neural network applied to defect identification of grounding grid,and provide reference for intelligent algorithm embedded in transient electromagnetic detection system.

关 键 词：接地网缺陷 瞬变场参数 视电阻率成像 ELMAN神经网络 改进鲸鱼算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] TM862[自动化与计算机技术—计算机科学与技术] TP183[电气工程—高电压与绝缘技术]