基于遗传算法优化的BP神经网络在密度界面反演中的应用

doi:10.13278/j.cnki.jjuese.201702301

摘要/Abstract

摘要： BP神经网络方法在二维密度界面的反演中取得了较好的效果，但在反演三维界面时，由于模型更复杂、参数更多，BP神经网络的收敛速度和反演精度都有一定程度的下降。为了改善反演效果，本文利用遗传算法对BP神经网络的权值、阈值选择过程进行优化，获得了更好的网络模型；并将此模型应用于密度界面模型的反演中，预测误差从上百米减小到数十米，同时迭代计算步数减少了近2/3，有效减少了计算时间，反演结果更准确。利用基于遗传算法优化的BP神经网络反演了法国某地区莫霍面深度，预测相对误差仅为1.8%，取得了较好的应用效果。基于遗传算法优化的BP神经网络在密度界面的反演中具有良好的应用价值和研究前景。

关键词: BP神经网络, 遗传算法, 密度界面反演, 网络训练, 优化

Abstract: The method of BP neural network has achieved good results in the inversion of 2D density interface, however, the converging speed and inversion accuracy would decrease when it is used to inverse 3D density interface due to more complicated model and more parameters. Genetic algorithm is used to optimize the process of choosing weights and thresholds of BP network in this paper in order to improve inversion results. Then a better network model is obtained and this model will be utilized in the inversion of density interface model. This method could increase inversion accuracy as well as reduce calculation time, and better inversion results would be achieved. At last the method is utilized to inverse the depth of Moho in some region in France and the application effect is good. It is illustrated that BP neural network based on genetic algorithm has benign application value and research prospect in the inversion of density interface.

Key words: BP neural network, genetic algorithm, inversion of density interface, network training, optimization

中图分类号:

P312.1

张代磊, 黄大年, 张冲. 基于遗传算法优化的BP神经网络在密度界面反演中的应用[J]. 吉林大学学报(地球科学版), 2017, 47(2): 580-588.

Zhang Dailei, Huang Danian, Zhang Chong. Application of BP Neural Network Based on Genetic Algorithm in the Inversion of Density Interface[J]. Journal of Jilin University(Earth Science Edition), 2017, 47(2): 580-588.

参考文献

[1] Van der Bann M, Jutten C. Neural Networks in Geo-physical Applications[J]. Society of Exploration Geophysists, 2000, 65(4):1032-1047.
[2] Russell B. Neural Network Applications in Geophysics[C]//CSEG National Convention. Cakgary:Hampon-Russell Software, 2005:339-341.
[3] Hajian A, Ardestani E V, Lucas C. Depth Estimation of Gravity Anomalies Using Hopfield Neural Networks[J]. Journal of the Earth & Space Physics, 2011, 37(2):1-9.
[4] Nagendra R, Prasad P V S, Bhimasankaram V L S. Forword and Inverse Computer Modeling of a Gravity Field Resulting from a Density Interface Using Parker-Oldenberg Method[J]. Computer & Sciences, 1996, 22(3):227-237.
[5] 陈东敬,张新兵. 带模拟退火的拟BP神经网络在伊朗某地区重力资料反演中的应用[J]. 勘探地球物理进展,2005, 28(3):215-218. Chen Dongjing, Zhang Xinbing. The Application of Quasi-BP Neural Network with Simulated Annealing in the Inversion of Gravity Data in Some Region of Iran[J]. Progress in Exploration Geophysics, 2005, 28(3):215-218.
[6] 朱自强,程方道,黄国祥. 同时反演两个三维密度界面的拟神经网络BP算法[J]. 石油物探,1995, 34(1):76-85. Zhu Ziqiang, Cheng Fangdao, Huang Guoxiang. Quasi Neural Network BP Algorithm for Simultaneous Inversion of 3-D Density Interface[J]. Geophysical Prospecting for Petroleum, 1995, 34(1):76-85.
[7] 张新兵,王家林,陈冰,等. BP, Hopfield神经网络在位场反演中的应用比较[J]. 物探化探计算技术,2007, 29(2):162-166. Zhang Xinbing, Wang Jialin, Chen Bing, et al. The Application Comparison of BP, Hopfield Neural Network in the Inversion of Potential Field[J]. Computing Techniques for Geophysical and Geochemical Exploration, 2007, 29(2):162-166.
[8] 郭文斌,朱自强,鲁光银. 重力异常的BP神经网络三维物性反演[J]. 地球物理学进展,2012, 27(2):409-416. Guo Wenbin, Zhu Ziqiang, Lu Guangyin. 3-D Gravity Inversion for Physical Properties Using BP Network[J]. Progress in Geophysics, 2012, 27(2):409-416.
[9] 侯祥林,陈长征,虞和济,等. 神经网络权值和阈值的优化方法[J]. 东北大学学报(自然科学版),1999, 20(4):447-450. Hou Xianglin, Chen Changzheng, Yu Heji, et al. Optimum Method About Weights and Thresholds of Neural Network[J]. Journal of Northeastern University (Natural Science), 1999, 20(4):447-450.
[10] 张国翊,胡铮. 改进BP神经网络模型及其稳定性分析[J]. 中南大学学报(自然科学版),2011, 42(1):115-124. Zhang Guoyi, Hu Zheng. Improved BP Neural Network Model and Its Stability Analysis[J]. Journal of Central South University (Science and Technology), 2011, 42(1):115-124.
[11] 刘洪林,胡建,王金山,等. GA+BP神经网络在薄互层储层预测中的应用[J]. 物探与化探,2004, 28(5):460-463. Liu Honglin, Hu Jian, Wang Jinshan, et al. The Application of GA+BP Hybrid Neural Network to Inter-Bed Deposit Prediction[J]. Geophysical & Geochemical Exploration, 2004, 28(5):460-463.
[12] 李琼,李勇,李正文,等. 基于GA-BP理论的储层视裂缝密度地震非线性反演方法[J]. 地球物理学进展,2006, 26(2):465-471. Li Qiong, Li Yong, Li Zhengwen, et al. A Seismic Nonlinear Inversion for Apparent Fracture Density of Hydrocarbon Reservior Based on GA-BP Theory[J]. Progress in Geophysics, 2006, 26(2):465-471.
[13] 柳建新,童孝忠,李爱勇,等. MT资料反演的一种实数编码混合遗传算法[J]. 中南大学学报(自然科学版),2007, 38(1):160-163. Liu Jianxin, Tong Xiaozhong, Li Aiyong, et al. A Real Coded Hybrid Genetic Algorithm in Magnetotelluric Sounding Data Inversion[J]. Journal of Central South University (Science and Technology), 2007, 38(1):160-163.
[14] 柯小平,王勇,许泽厚,等. 青藏东缘三维Moho界面的位场遗传算法反演[J]. 大地测量与地球动力学,2006, 26(1):101-104. Ke Xiaoping, Wang Yong, Xu Zehou, et al. 3D Moho Depth Inversion of Eastern Tibetan Plateau from Gravity Data with Genetic Algorithm[J]. Journal of Geodesy and Geodynamics, 2006, 26(1):101-104.
[15] 徐黎明,王清,陈剑平,等. 基于BP神经网络的泥石流平均流速预测[J]. 吉林大学学报(地球科学版), 2013, 43(1):186-191. Xu Liming, Wang Qing, Chen Jianping, et al. Forecast for Average Velocity of Debris Flow Based on BP Neural Network[J]. Journal of Jilin University (Earth Science Edition), 2013, 43(1):186-191.
[16] 张明君,张化光. 基于遗传算法优化的神经网络PID控制器[J]. 吉林大学学报(工学版),2005, 35(1):92-96. Zhang Mingjun, Zhang Huaguang. Neural Network PID Controller Optimized by GA[J]. Journal of Jilin University (Engineering and Technology Edition), 2005, 35(1):92-96.
[17] 王贝贝,郝天珧. 具有已知深度点的二维单一密度界面的反演[J]. 地球物理学进展,2008, 23(3):834-838. Wang Beibei, Hao Tianyao. The Inversion of Two-Dimensional Mono-Density Interface with Several Known Control Points[J]. Progress in Geophysics, 2008, 23(3):834-838.
[18] 胡立天,郝天珧. 带控制点的三维密度界面反演方法[J]. 地球物理学进展,2014, 29(6):2498-2503. Hu Litian, Hao Tianyao. The Inversion of Three-Dimensional Density Interface with Control Points[J]. Progress in Geophysics, 2014, 29(6):2498-2503.
[19] 刘银萍,孟令顺. 利用重力异常研究虎林盆地的构造分区和基底形态[J]. 吉林大学学报(地球科学版),2007, 37(增刊1):36-39. Liu Yinping, Meng Lingshun. Using Gravity Anomaly Study Structure Division and Basement Morphology of the Hulin Basin[J]. Journal of Jilin University (Earth Science Edition), 2007, 37(Sup. 1):36-39.
[20] 焦新华,陈化然,吴燕冈,等. 天津地区地壳物性界面的计算及深部构造特征[J]. 吉林大学学报(地球科学版),2006, 36(4):616-621. Jiao Xinhua, Chen Huaran, Wu Yangang, et al. The Calculation and Study of Petrophysical Property Interfaces and Their Deep Structural Feature in Tianjin Area[J]. Journal of Jilin University (Earth Science Edition), 2006, 36(4):616-621.

相关文章 15

[1]	肖凡, 陈建国. 基于RCGA的PPC模型在化探异常识别与提取中的应用[J]. 吉林大学学报(地球科学版), 2017, 47(4): 1319-1330.
[2]	卢文喜, 郭家园, 董海彪, 张宇, 林琳. 改进的支持向量机方法在矿山地质环境质量评价中的应用[J]. 吉林大学学报(地球科学版), 2016, 46(5): 1511-1519.
[3]	吴鸣, 吴剑锋, 施小清, 刘杰, 陈干, 吴吉春. 基于谐振子遗传算法的高效地下水优化管理模型[J]. 吉林大学学报(地球科学版), 2015, 45(5): 1485-1492.
[4]	余楚, 张翼龙, 孟瑞芳, 曹文庚. 河套平原浅层地下水动态监测网优化设计[J]. 吉林大学学报(地球科学版), 2015, 45(4): 1173-1179.
[5]	王宇, 卢文喜, 卞建民, 侯泽宇. 三种地下水位动态预测模型在吉林西部的应用与对比[J]. 吉林大学学报(地球科学版), 2015, 45(3): 886-891.
[6]	杜润林, 刘展. 基于粒子群优化的细胞神经网络油气重力异常信息提取[J]. 吉林大学学报(地球科学版), 2015, 45(3): 926-933.
[7]	刘贺，张弘强，刘斌. 基于粒子群优化神经网络算法的深基坑变形预测方法[J]. 吉林大学学报(地球科学版), 2014, 44(5): 1609-1614.
[8]	王洪德，高幼龙，薛星桥，金枭豪，王刚. 典型滑坡监测点优化布置[J]. 吉林大学学报(地球科学版), 2013, 43(3): 858-866.
[9]	彭帅英，李广杰，彭文，马建全，王雪冬，秦胜伍. 基于改进遗传算法的Holt-Winters模型在采空沉陷预测中的应用[J]. 吉林大学学报(地球科学版), 2013, 43(2): 515-520.
[10]	秦宁，李振春，杨晓东，张凯，王俊. 共散射点道集与角道集串级优化叠前偏移速度分析[J]. 吉林大学学报(地球科学版), 2013, 43(2): 623-631.
[11]	徐黎明，王清,陈剑平，潘玉珍. 基于BP神经网络的泥石流平均流速预测[J]. 吉林大学学报(地球科学版), 2013, 43(1): 186-191.
[12]	周福军，陈剑平，栾海，徐黎明，牛岑岑. 变尺度混沌优化算法在二密滑坡锚固方案优化设计中的应用[J]. 吉林大学学报(地球科学版), 2013, 43(1): 192-198.
[13]	李鸿雁, 赵娟, 王玉新, 韩振, 王傲. 扩域搜索遗传算法优化马斯京根参数及其应用[J]. J4, 2011, 41(3): 861-865.
[14]	周晓华, 林君, 陈祖斌, 焦健, 郭同健. 改进的神经网络反演微动面波频散曲线[J]. J4, 2011, 41(3): 900-906.
[15]	温忠辉, 任化准, 束龙仓, 王恩, 柯婷婷, 陈荣波. 岩溶地下河日流量预测的小样本非线性时间序列模型[J]. J4, 2011, 41(2): 455-458.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed