SOM-RBF神经网络模型在地下水位预测中的应用应用

doi:10.13278/j.cnki.jjuese.201501204

吉林大学学报(地球科学版) ›› 2015, Vol. 45 ›› Issue (1): 225-231.doi: 10.13278/j.cnki.jjuese.201501204

SOM-RBF神经网络模型在地下水位预测中的应用应用

刘博^1,2, 肖长来^1,2, 梁秀娟^1,2

1. 吉林大学环境与资源学院, 长春 130021;
2. 吉林大学地下水资源与环境教育部重点实验室, 长春 130021

收稿日期:2014-06-26 发布日期:2015-01-26
通讯作者: 肖长来(1962), 男, 教授, 博士生导师, 主要从事资源与水环境、地下水资源评价与开发利用研究, E-mail:xcl2822@126.com E-mail:xcl2822@126.com
作者简介:刘博(1987), 男, 博士研究生, 主要从事地下水科学与工程研究, E-mail:liu-bo727@163.com
基金资助:
吉林省科技引导项目(20080543);高等学校博士学科点专项科研基金项目(200801830044);教育部国家潜在油气资源项目(OSR-01-07)

Application of Combining SOM and RBF Neural Network Model for Groundwater Levels Prediction

Liu Bo^1,2, Xiao Changlai^1,2, Liang Xiujuan^1,2

1. College of Environment and Resources, Jilin University, Changchun 130021, China;
2. Key Laboratory of Groundwater Resources and Environment, Ministry of Education, Jilin University, Changchun 130021, China

Received:2014-06-26 Published:2015-01-26

摘要/Abstract

摘要：

利用自组织映射(SOM) 聚类模型优化径向基函数神经网络(RBFN)隐层节点的方法, 减小了RBFN由于自身结构问题在地下水水位预测中产生的误差。采用SOM对已有样本进行聚类, 利用聚类后的二维分布图确定隐层节点的数目, 并根据聚类结果计算径向基函数的宽度, 确定径向基函数的中心, 由此建立SOM-RBFN模型。以吉林市丰满区二道乡为例, 采用20002009年观测的地下水位动态资料, 利用SOM-RBFN模型对地下水位进行预测, 验证其准确性, 并分别以5、7、10 a的地下水位动态数据为研究样本建立模型, 考查样本数量对预测结果的影响。研究结果表明:SOM-RBFN模型预测地下水水位过程中, 均方根误差(RMSE)的均值为0.43, 有效系数(CE)的均值为0.52, 均达到较高标准, 因此SOM-RBFN模型可以作为有效而准确的地下水水位预测方法;同时RBF7的RMSE和CE均值分别为0.38和0.68, 结果优于RBF5和RBF10, 这就意味着在模型计算中样本数量不会直接影响预测结果的精度。

关键词: 地下水位预测, SOM, RBF, 神经网络

Abstract:

As the hidden units of radial basis function network (RBF) were optimized by the theory of self-organizing map (SOM), the groundwater levels forecasting error range, due to its structural problems, could be reduced. With the two-dimensional feature map and clustering results of SOM, the number of hidden units, the position and the width of the radial basis centers can be easily determined. The SOM-RBFN model can be established. The accuracy of the model was verified by predicting groundwater level at Erdao Town in Fengman District of Jilin City based on observed groundwater level from 2000 to 2009.In addition, dynamic data of groundwater level for five years (2005-2009), seven years (2003-2009), ten years (2000-2009), are used as study samples and make forecast one by one, which can examine that if the sample size could influence the forecast result. The results prove that SOM-RBFN model can be used in groundwater levels dynamic forecasting, because the averages of RMSE and CE are 0.43 and 0.52, respectively, which are the relatively good outcomes. And, the averages of RMSE and CE of RBF7 are 0.38 and 0.68, whose results are better than RBF5 and RBF10. Therefore, it can be known that the amount of data cannot directly influence the accuracy of results.

Key words: groundwater level prediction, self-organizing map, radial basis function, neural networks

中图分类号:

P641

刘博, 肖长来, 梁秀娟. SOM-RBF神经网络模型在地下水位预测中的应用应用[J]. 吉林大学学报(地球科学版), 2015, 45(1): 225-231.

Liu Bo, Xiao Changlai, Liang Xiujuan. Application of Combining SOM and RBF Neural Network Model for Groundwater Levels Prediction[J]. Journal of Jilin University(Earth Science Edition), 2015, 45(1): 225-231.

参考文献

[1] 金菊良, 杨晓华, 金宝明, 等.基于遗传算法的地下水位动态预测双线性模型[J].水科学进展, 2001, 12(3):361-365. Jin Juliang, Yang Xiaohua, Jin Baoming, et al. A Bilinear Time Series Model Based on Genetic Algorithm for Predicting Groundwater Level Regime[J]. Advances in Water Science, 2001, 12(3):361-365.

[2] Savic D A, Walters G A, Davidson J W. A Genetic Programming Approach to Rainfall-Runoff Modeling[J]. Water Resour Manage, 1999, 13(3):219-231.

[3] Chen S K, Mangimeli P, West D. The Comparative Ability of Self-Organizing Neural Networks to Define Cluster Structure[J]. Omega, 1995, 23(3):271-279.

[4] 杜超, 肖长来, 吕军, 等.地下水水质实时预报系统开发及利用:以下辽河平原为例[J].吉林大学学报:地球科学版, 2014, 44(5):1625-1632. Du Chao, Xiao Changlai, Lü Jun, et al. Development and Application of Real Time Groundwater Quality Prediction System:An Example in Lower Liaohe River Plain[J]. Journal of Jilin University:Earth Science Edition, 2014, 44(5):1625-1632.

[5] Box G E P, Jenkins G M. Time Series Analysis, Forecasting and Control[M]. San Francisco:Holden-Day, 1976.

[6] Hipel K W, Mcleod A I. Time Series Modeling of Water Resources and Environmental Systems[M]. Amsterdam:Elsevier, 1994.

[7] Knotters M, Van Walsum V. Estimation Fluctuation Quantities from Time Series of Water Table Depths Using Models with a Stochastic Component[J]. Journal of Hydrology, 1997, 197(1/2/3/4):25-46.

[8] Ahn H, Salas J D. Groundwater Head Sampling Based on Stochastic Analysis[J]. Water Resour Res, 1997, 33(12):2769-2780.

[9] Kumar D, Ahmed S. Seasonal Behaviour of Spatial Variability of Groundwater Level in a Granitic Aquifer in Monsoon Climate[J]. Current Science, 2003, 84(25):188-196.

[10] Gundogdu K S, Guney I.Spatial Analyses of Ground-water Levels Using Universal Kriging[J]. Journal of Earth System Science, 2007, 116(1):49-55.

[11] Nag A K, Mitra A. Forecasting the Dailu Foreign Exchange Rates Using Genetically Optimized Neural Networks[J]. Journal of Forecasting, 2002, 21(7):501-511.

[12] Zhang G, Patuwo B E, Hu M Y. Forecasting with Artificial Neural Networks:The State of the Art[J]. International Journal of Forecasting, 1998, 14(1):35-60.

[13] Bulsari A, Saxen H. Nonlinear Time Series Analysis by Neural Networks:A Case Study[J]. Int J Neural Syst, 1993, 12(1):1-10.

[14] Atiya A F. Prediction for Credit Risk Using Neural Networks:A Survey and New Results[J]. IEEE Trans Neural Netw, 2001, 12(4):929-935.

[15] 孟艳, 潘宏侠.PSO聚类和梯度算法结合的RBF神经网络优化[J].自动化仪表, 2011, 32(2):6-8. Meng Yan, Pan Hongxia. Optimization of RBF Neural Network Based on Combination of PSO Clustering and Gradient Algorithms[J]. Process Automation Insterumentation, 2011, 32(2):6-8.

[16] 朱明星, 张德龙. RBF网络基函数中心选取的算法研究[J]. 安徽大学学报:自然科学版, 2000, 24(1):72-74. Zhu Mingxing, Zhang Delong. Study on the Algorithms of Selecting the Radial Basis Function Center[J]. Journal of Anhui University:National Science Edition, 2000, 24(1):72-74.

[17] 方力智, 张翠芳, 易芳.基于改进差分进化算法的RBF神经网络优化方法[J].成都大学学报:自然科学版, 2009, 28(3):231-234. Fang Lizhi, Zhang Cuifang, Yi Fang. Optimization Approach Based on Modified Differential Evolution Algorithm for RBF Neural Network[J]. Journal of Chengdu University:Natural Science Edition, 2009, 28(3):231-234.

[18] 张江涛, 刘旭敏.一种基于代数算法的RBF神经网络优化方法[J].计算机工程与应用, 2008, 44(5):96-98. Zhang Jiangtao, Liu Xumin. Optimization Approach Based on Algebraic Algorithm for RBF Neural Network[J]. Computer Engineering and App-lications, 2008, 44(5):96-98.

[19] 马洪伟, 赵志刚, 吕慧显, 等.基于蚁群聚类和裁剪方法的RBF神经网络优化算法[J].青岛大学学报:工程技术版, 2008, 23(3):35-38. Ma Hongwei, Zhao Zhigang, Lü Huixian, et al. Optimization Algorithm of RBF Neural Networks Based on Ant Colony Clustering and Pruning Method[J]. Journal of Qingdao University:Engineering and Technology, 2008, 23(3):35-38.

[20] 涂松, 贲可荣, 徐荣武, 等.采用SOM和RBF优化下的水下航行器噪声源识别[J]. 计算机工程与科学, 2007, 29(10):112-114. Tu Song, Ben Kerong, Xu Rongwu, et al. Identification of the Acoustic Fault Sources of the Underwater Vehicles Based on the SOM and RBF Neural Network Optimization[J]. Computer Engineering & Science, 2007, 29(10):112-114.

[21] Chen Luhsienm, Chen Chingtien, Pan Yangu.Ground-water Level Prediction Using SOM-RBFN Multisite Model[J]. Journal of Hydrologic Engineering, 2010, 15(8):624-625.

[22] ASCE Task Committee on Application of Artificial Neural Networks in Hydrology. Artificial Neural Network in Hydrology[J]. Journal of Hydrologic Eng, 2000, 5(2):142-137.

[23] Haykin S.Neural Networks:A Comprehensive Foun-dation[M]. London:Macmillan, 1994.

Metrics

Viewed

Full text

Abstract

Cited

Shared

Discussed

SOM-RBF神经网络模型在地下水位预测中的应用应用

Application of Combining SOM and RBF Neural Network Model for Groundwater Levels Prediction

PDF (PC)

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 15

Metrics

本文评价

推荐阅读 0

[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.
[2]	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.
[3]	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.
[4]	潘保芝, 石玉江, 蒋必辞, 刘丹, 张海涛, 郭宇航, 杨小明. 致密砂岩气层压裂产能及等级预测方法[J]. 吉林大学学报(地球科学版), 2015, 45(2): 649-654.
[5]	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.
[6]	王羽,许强,柴贺军,刘丽,夏毓超,王晓东. 工程岩爆灾害判别的RBF-AR耦合模型[J]. 吉林大学学报(地球科学版), 2013, 43(6): 1943-1949.
[7]	鲁功达,晏鄂川,王环玲,王雪明,谢良甫. 基于岩石地质本质性的碳酸盐岩单轴抗压强度预测[J]. 吉林大学学报(地球科学版), 2013, 43(6): 1915-1921.
[8]	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.
[9]	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.
[10]	朴金石, 殷琨, 范黎明. 利用神经网络法预测风动潜孔锤钻速[J]. J4, 2009, 39(5): 882-886.
[11]	张晨，陈剑平,肖云华. 基于神经网络对有限元强度折减法分析[J]. J4, 2009, 39(1): 114-0118.
[12]	郄瑞卿，薛林福，王满，王丽华. SOFM储层综合评价方法及其在延吉盆地的应用[J]. J4, 2009, 39(1): 168-0174.
[13]	秦胜伍，陈剑平. 隧道围岩压力的神经网络时间序列分析[J]. J4, 2008, 38(6): 1005-1009.
[14]	郝立波，蒋艳明，陆继龙，孙淑梅，白荣杰. 利用多目标地球化学数据识别第四纪沉积物类型--基于概率神经网络方法[J]. J4, 2008, 38(6): 1081-1084.
[15]	邱道宏，陈剑平，阙金声，安鹏程. 基于粗糙集和人工神经网络的洞室岩体质量评价[J]. J4, 2008, 38(1): 86-0091.