电工技术学报  2017, Vol. 32 Issue (12): 146-154    DOI:
电工理论与新技术 |
遗传神经网络的瞬变电磁视电阻率求解算法
秦善强1, 付志红1, 朱学贵1, 籍勇亮2
1. 输配电装备及系统安全与新技术国家重点实验室(重庆大学) 重庆 400044;
2. 国网重庆电力公司电力科学研究院 重庆 401123
Genetic Neural Network for Apparent Resistivity Solution of Transient Electromagnetic
Qin Shanqiang1, Fu Zhihong1, Zhu Xuegui1, Ji Yongliang2
1. State Key Laboratory of Power Transmission Equipment & System Security and New Technology Chongqing University Chongqing 400044 China;
2. Electric Power Research Institute Chongqing Electric Power Company Chongqing 401123 China
全文: PDF (25267 KB)  
输出: BibTeX | EndNote (RIS)      
摘要 提出用遗传神经网络求解中心回线装置下瞬变电磁法(TEM)的视电阻率。根据中心回线方式的瞬变电磁响应关系式,设计出神经网络的输入输出关系和单输入单输出的三层网络结构。计算出神经网络的输入输出样本集,并通过尝试法确定隐含层的神经元个数。引入遗传算法优化神经网络结构的连接权值,得到了最优连接权值的GABP神经网络。用该神经网络对瞬变电磁响应的非线性方程进行拟合,得到以实测数据计算的核函数值所一一对应的瞬变场参数值,达到求视电阻率并成像的目的。通过对地下高阻块状异常体模型和电力系统钢制扁钢材料的接地网模型两个实例模型的仿真计算证明,得到视电阻率断面图,达到了求解反问题的效果。理论模型与实际数据计算表明,遗传优化的BP神经网络使得瞬变电磁视电阻率的计算时间大大缩短,是个实用的算法。该方法为瞬变电磁接地网故障实时诊断平台提供必要的技术基础。
服务
把本文推荐给朋友
加入我的书架
加入引用管理器
E-mail Alert
RSS
作者相关文章
秦善强
付志红
朱学贵
籍勇亮
关键词 瞬变电磁场 人工神经网络 遗传算法 反向传播 视电阻率    
Abstract:The genetic neural network is applied to compute the central-loop transient electromagnetic (TEM) apparent resistivity. According to the central loop TEM response, a relationship between inputs and outputs of an ANN as well as three-layer architecture with single input and single output for ANN is designed. The ANN sample sets are calculated, and the number of hidden layer neuron is determined through trial method. A genetic algorithm is introduced to optimize connection weights of the ANN and then a GABP neural network with the optimal connection weights. A nonlinear equation of the TEM response is fitted by GABP neural network, and transient parameter value is obtained. The transient parameter value and kernel value from measured data are corresponding one by one. Finally, the apparent resistivity is calculated. Two examples with GABP neural network are presented. The first one is a forward calculated model of high impedance abnormal body in uniform half space, the second one is the grounding grid with the structural shape ‘?’ in a substation of power system. Pseudo-section of apparent resistivity is obtained, and good effects to solve inverse problem is achieved. The comparisons between the theoretical models and the measured data show that the GABP is a usefulness algorithm, which can reduce much computing time of TEM apparent resistivity. This method provides the base technology for a real-time fault diagnosis platform of grounding grids using TEM.
Key wordsTransient electromagnetic field    artificial neural networks    genetic algorithm    backpropagation    apparent resistivity   
收稿日期: 2015-10-14      出版日期: 2017-06-30
PACS: TM153.1  
基金资助:国家自然科学基金项目(51277189),输配电装备及系统安全与新技术国家重点实验室自主研究重点项目(2007DA10512714103)及国家电网公司总部科技项目(基于瞬变电磁法的接地网状态检测及故障诊断新技术研究)资助
通讯作者: 秦善强 男,1982年生,博士研究生,主要研究方向为电磁探测与成像技术,多层感知器前馈神经网络和遗传算法等。E-mail: qinshanqiang112@163.com   
作者简介: 付志红 男,1966年生,博士,教授,博士生导师,主要研究方向为电能质量分析与电能计量、电力电子、电磁探测技术等。E-mail: fuzhihong@cqu.edu.cn
引用本文:   
秦善强, 付志红, 朱学贵, 籍勇亮. 遗传神经网络的瞬变电磁视电阻率求解算法[J]. 电工技术学报, 2017, 32(12): 146-154. Qin Shanqiang, Fu Zhihong, Zhu Xuegui, Ji Yongliang. Genetic Neural Network for Apparent Resistivity Solution of Transient Electromagnetic. Transactions of China Electrotechnical Society, 2017, 32(12): 146-154.
链接本文:  
https://dgjsxb.ces-transaction.com/CN/Y2017/V32/I12/146