基于人工神经网络模型的黑河流域径流模拟预报
|
摘要
水资源是基础性的自然资源,是维系自然界一切生命系统不可缺少的资源,也是社会经济发展所必需的不可替代的资源。黑河流域位于西北内陆干旱地区,近年来黑河流域出现了一系列的生态环境问题,水问题是引起这一系列问题的关键问题。为了解决黑河流域存在的社会﹑经济﹑生态环境问题,发挥黑河流域水资源的最大经济﹑环境和社会效益,使黑河流域的社会﹑经济﹑生态环境得到可持续发展,必须对黑河流域的水资源系统进行准确而及时的分析与预测,合理配置,优化调度,积极有效保护有限的黑河流域水资源,最大限度的满足黑河流域国民经济各部门的合理需求,促进资源与环境,生态系统的良性循环,以及经济﹑社会﹑环境协调发展。要做到这些,必须对黑河上游流域水文水资源及其变化特性有一个比较清晰和全面的认识,因此,预报黑河水资源量的时空分布和水资源量的多少对于黑河流域社会经济的发展、生态环境的保护和水资源管理具有十分重要的意义。
人工神经网络是一种数据驱动模型,它可以客观地描述系统内部的真正构成,不受以往知识的束缚,从而找到系统输入输出之间的合理的关系,找出内部的规律性,进行精确的模拟,并为径流预报提供辅助手段,但这也并不是说要建立一个纯黑箱的模型,在建立径流模拟和预报的人工神经网络模型时,应该把流域的物理背景知识考虑在内。
莺落峡和正义峡是黑河干流上的两个重要控制站,首先对影响莺落峡,正义峡径流变化的因素进行分析,得知莺落峡径流量变化受人为因素的影响比较小,主要受自然因素的影响,正义峡径流量受人为因素影响比较大;然后考虑流域的物理背景,建立BP人工神经网络模型,对莺落峡月径流量的模拟主要考虑自然因素。单一构型的BP人工神经网络模型是一种顾及各种数据情况的所谓的“最佳协调解”。这种模型无法识别各个径流阶段的主要影响因素并分别对待,高流量、低流量等事件相关的个性信息没有表现的空间,因而一般会被当作噪声过滤掉。这样把所有可得数据混在一起训练出的神经网络,往往只可以较好地模拟一般状况(如中等流量事件,这种事件的样本数目较多),而对极值状况,模型的性能则较差,无法使各个径流时期的径流模拟同时达到令人满意的程度,建立莺落峡月径流模拟模型的时候,采用模糊聚类的方法对径流进行分类,分为汛期和非汛期,然后分别建立BP人工神经网络模型,对比分类和未分类的BP模型模拟结果,发现分类之后所建立的模型地性能比较好,因此采用分类之后的模型对莺落峡的月径流进行了预测。
正义峡的年径流量模拟模型的建立分别考虑人为因素和不考虑人为因素,建立两个模型,并做出比较分析,考虑人为因素的模型的确定性系数在检验期和训练期都高于不考虑人为因素的模型。考虑人为因素之后的BP模型要比不考虑人为因素的模型要好,说明在建立模型的时候,应该结合实际情况,不仅要考虑自然因素,对受人类影响较大的径流的模拟时,应该考虑物理背景,这样才能提高模型的性能。
建立莺落峡和正义峡的径流模拟模型的目的是为了更好的预测,选取模拟精度比较高的模型进行预测,在不同的气候情景下,利用BP网络模拟计算2030年黑河流域莺落峡水文站的径流量,在降水量不变的情况下,气温增加0.5℃,年径流量将增加8.92 %;如气温增加1℃,则年径流量将减少5.414 %;保持气温不变,降水量增加10 %,年径流量将增加9.905 %;如气温增加0.5℃,同时降水量增加10 %,则年径流量将增加8.98 %。
在全球变化的背景下,黑河流域正义峡的径流量,在2030年之前,年径流量将有一定程度的增加,但幅度不会很大,以后随着气温的继续升高,正义峡的径流将会减少,该结果与康尔泗等(1999)研究结果较为一致。
Water resource is an essential natural resource, it is indispensable to living system and can not be replaced by other resource. Heihe river basin locates in arid region of northwest. In recent years, a series of ecological problems have occurred in Heihe river basin. Water problem is the key problem that causes those ecological problems. In order to solve the social﹑ecological and economic problem, lets the social﹑economic and environment be sustainable, it is necessary to precisely analyze and predicate the water resource system, rational allocate, optimal operation and effectively protect limited water resource of Heihe river basin, maximum meet the need of each national economic department, promote the economic﹑social and environment to harmoniously develop. In order to do those, it is need to know the water resource clearly and comprehensively, so it has a very important meaning to social-economic development, ecological environment protection and water resource management and protection to predicate the spatial distribution and quantity.
Artificial neural network is a data-driven model, It can objectively describe the real constitution of system, ANN model is not affected by previous knowledge, it can find the reasonable relation between input and output, find the inner rule and do precisely simulate, so ANN model can be a supplementary means to runoff predication. This does not say that it build a pure black-box model, when build an ANN model to simulate and predicate runoff, we should consider the physics knowledge of river basin.
Yingluoxia and Zhengyixia are two key control station of mainstream in Heihe, in order to predicate the runoff of Heihe, Firstly, analyze these factors that affect the change of runoff in Yingluoxia and Zhengyixia, find that the runoff of yingluoxia is mainly affected by natural factors, while the runoff of Zhengyixia is mainly affected by human factors; Secondly, consider the physical background of river basin, build back propagation ANN model, to simulate the monthly runoff of Yingluoxia, we mainly consider natural factors. The BP ANN with unitary configuration is a optimal coordinate solution, so it can not discern the main affecting factor in each stage and can not separately treat, the maximum and minimum runoff is often neglected, so this type ANN model can often simulate the median value very well, but to the extreme value, the performance of model is bad, so when we build model of the monthly runoff in Yingluoxia, we classify the runoff using the fuzzy cluster method, divided the runoff into flood season and non-flood season, separately build ANN model. Compare the performance of BP ANN model that consider the classification and non-classification, find the model with classification has better performance, so we use this model simulate the runoff of Yingluoxia.
When build BP ANN model of annual runoff in Zhengyixia, in order to consider the physical background, we separately build the ANN model, one considering the human factor, the other not, after training, analyze the performance of two models, finds that deterministic coefficients of the model considering the human factor in training stage and test stage are higher than the model not considering the human factor, so in the future research, the physical background should be considered, so the performance of ANN model can improve.
The aim of build the simulated ANN model of Yingluoxia and Zhengyixia is to predicate the runoff, so we choose the model that has high performance to predicate the runoff in different climate background, the results are good.
人工神经网络是一种数据驱动模型,它可以客观地描述系统内部的真正构成,不受以往知识的束缚,从而找到系统输入输出之间的合理的关系,找出内部的规律性,进行精确的模拟,并为径流预报提供辅助手段,但这也并不是说要建立一个纯黑箱的模型,在建立径流模拟和预报的人工神经网络模型时,应该把流域的物理背景知识考虑在内。
莺落峡和正义峡是黑河干流上的两个重要控制站,首先对影响莺落峡,正义峡径流变化的因素进行分析,得知莺落峡径流量变化受人为因素的影响比较小,主要受自然因素的影响,正义峡径流量受人为因素影响比较大;然后考虑流域的物理背景,建立BP人工神经网络模型,对莺落峡月径流量的模拟主要考虑自然因素。单一构型的BP人工神经网络模型是一种顾及各种数据情况的所谓的“最佳协调解”。这种模型无法识别各个径流阶段的主要影响因素并分别对待,高流量、低流量等事件相关的个性信息没有表现的空间,因而一般会被当作噪声过滤掉。这样把所有可得数据混在一起训练出的神经网络,往往只可以较好地模拟一般状况(如中等流量事件,这种事件的样本数目较多),而对极值状况,模型的性能则较差,无法使各个径流时期的径流模拟同时达到令人满意的程度,建立莺落峡月径流模拟模型的时候,采用模糊聚类的方法对径流进行分类,分为汛期和非汛期,然后分别建立BP人工神经网络模型,对比分类和未分类的BP模型模拟结果,发现分类之后所建立的模型地性能比较好,因此采用分类之后的模型对莺落峡的月径流进行了预测。
正义峡的年径流量模拟模型的建立分别考虑人为因素和不考虑人为因素,建立两个模型,并做出比较分析,考虑人为因素的模型的确定性系数在检验期和训练期都高于不考虑人为因素的模型。考虑人为因素之后的BP模型要比不考虑人为因素的模型要好,说明在建立模型的时候,应该结合实际情况,不仅要考虑自然因素,对受人类影响较大的径流的模拟时,应该考虑物理背景,这样才能提高模型的性能。
建立莺落峡和正义峡的径流模拟模型的目的是为了更好的预测,选取模拟精度比较高的模型进行预测,在不同的气候情景下,利用BP网络模拟计算2030年黑河流域莺落峡水文站的径流量,在降水量不变的情况下,气温增加0.5℃,年径流量将增加8.92 %;如气温增加1℃,则年径流量将减少5.414 %;保持气温不变,降水量增加10 %,年径流量将增加9.905 %;如气温增加0.5℃,同时降水量增加10 %,则年径流量将增加8.98 %。
在全球变化的背景下,黑河流域正义峡的径流量,在2030年之前,年径流量将有一定程度的增加,但幅度不会很大,以后随着气温的继续升高,正义峡的径流将会减少,该结果与康尔泗等(1999)研究结果较为一致。
Water resource is an essential natural resource, it is indispensable to living system and can not be replaced by other resource. Heihe river basin locates in arid region of northwest. In recent years, a series of ecological problems have occurred in Heihe river basin. Water problem is the key problem that causes those ecological problems. In order to solve the social﹑ecological and economic problem, lets the social﹑economic and environment be sustainable, it is necessary to precisely analyze and predicate the water resource system, rational allocate, optimal operation and effectively protect limited water resource of Heihe river basin, maximum meet the need of each national economic department, promote the economic﹑social and environment to harmoniously develop. In order to do those, it is need to know the water resource clearly and comprehensively, so it has a very important meaning to social-economic development, ecological environment protection and water resource management and protection to predicate the spatial distribution and quantity.
Artificial neural network is a data-driven model, It can objectively describe the real constitution of system, ANN model is not affected by previous knowledge, it can find the reasonable relation between input and output, find the inner rule and do precisely simulate, so ANN model can be a supplementary means to runoff predication. This does not say that it build a pure black-box model, when build an ANN model to simulate and predicate runoff, we should consider the physics knowledge of river basin.
Yingluoxia and Zhengyixia are two key control station of mainstream in Heihe, in order to predicate the runoff of Heihe, Firstly, analyze these factors that affect the change of runoff in Yingluoxia and Zhengyixia, find that the runoff of yingluoxia is mainly affected by natural factors, while the runoff of Zhengyixia is mainly affected by human factors; Secondly, consider the physical background of river basin, build back propagation ANN model, to simulate the monthly runoff of Yingluoxia, we mainly consider natural factors. The BP ANN with unitary configuration is a optimal coordinate solution, so it can not discern the main affecting factor in each stage and can not separately treat, the maximum and minimum runoff is often neglected, so this type ANN model can often simulate the median value very well, but to the extreme value, the performance of model is bad, so when we build model of the monthly runoff in Yingluoxia, we classify the runoff using the fuzzy cluster method, divided the runoff into flood season and non-flood season, separately build ANN model. Compare the performance of BP ANN model that consider the classification and non-classification, find the model with classification has better performance, so we use this model simulate the runoff of Yingluoxia.
When build BP ANN model of annual runoff in Zhengyixia, in order to consider the physical background, we separately build the ANN model, one considering the human factor, the other not, after training, analyze the performance of two models, finds that deterministic coefficients of the model considering the human factor in training stage and test stage are higher than the model not considering the human factor, so in the future research, the physical background should be considered, so the performance of ANN model can improve.
The aim of build the simulated ANN model of Yingluoxia and Zhengyixia is to predicate the runoff, so we choose the model that has high performance to predicate the runoff in different climate background, the results are good.
引文
[1]王玲.基于人工神经网络的水文过程模拟研究[D].南京:河海大学,2002.
[2] Pankaj Singh, M.C.Deo. Suitability of different neural networks in daily flow orecasting[J]. Applied Soft Computing, 2007(7):968-978.
[3] Cameron M., Zealand, Donald H., etal. Short term streamflow forecasting using artificial neural networks[J]. Journal of Hydrology, 1999(214):32-48.
[4] Sanaga, Srinivasulu, Ashu Jain. A comparative analysis of training methods for artificial neural network rainfall–runoff models[J]. Applied Soft Computing, 2006(6):295-306.
[5] S.Riad, J.Mania. Rainfall-Runoff model using an artificial neural network approach[J]. Mathematical and Computer Modelling, 2004(40):839-846.
[6] Mahmut Firat, Mahmud Giiongor. River flow estimation using adaptive neuro fuzzy inference system[J]. Mathematics and Computers in Simulation, 2006(14):32-48.
[7] Wen Wang, Pieter H.A.J.M., Van Gelder. Forecasting daily streamflow using hybrid ANN models[J]. Journal of Hydrology, 2006(324):383–399.
[8] Drasko Furundzic. Application example of neural networks four time series analysis[J]. Rainfall-runoff modeling, 1998(64):383-396.
[9] A.Castelletti, R.Soncini-Sessa. Bayesian Networks and participatory masodeling in water resource management[J]. Environmental Modelling & Software, 2007(22):1075-1088.
[10] Y.B.Dibike, D.P.Solomatlne. River flow forecasting using artificial neural networks[J]. Phys. Chem. Earth(B), 2001,26(1):1-7.
[11]张立明.人工神经网络的模型及其应用[M].复旦大学出版社,1993.
[12]王文,马骏.若干水文预报方法综述[J].水利水电科技进展,2005,25(1):56-60.
[13]吴险峰,刘昌明.流域水文模型研究的若干进展[J].地理科学进展,2002,21(4):341-348.
[14] Famili A., Shen Weimin. Data preprocessing and intelligent data analysis,intelligent data analysis[Z].
[15] Feng C., Miehie D. Machine learning of rules and trees in machine learning[M]. Neural and statistical Classification, 1994.
[16]王栋,曹升乐.人工神经网络在水文水资源水环境系统中的应用研究进展[J].水利水电技术,1999,30(12):4-7.
[17]杨荣富,丁晶,刘国东.神经网络模拟降雨径流过程[J].水利学报,1998(10):69-73.
[18]陈仁升,康尔泗,张济世.应用GRNN神经网络模型计算西北干旱区内陆河流域出山径流[J].水科学进展,2002,13(1):87-91.
[19]徐中民,蓝永超,程国栋.人工神经网络方法在径流预报中的应用[J].冰川冻土,2000,22(4):372-375.
[20]王玲,黄国如.基于径流分类的日径流预测神经网络模型[J].灌溉排水,2002,21(4):45-48.
[21]王玲,黄国如.改进的BP网络模型在流于日径流量预测中的应用[J].水电能源科学,2003,21(1):32-34.
[22]张婧婧,姜铁兵,康玲等.基于人工神经网络的日径流预测[J].水自动化和大坝监测,2002,26(4):65-67.
[23]尹雄锐.基于智能计算的降雨径流模拟方法研究[D].武汉大学,2004.
[24]张少文,张学成,王玲等.黄河年降雨-径流BP预测模型研究[J].人民黄河,2005,27(1):18-20.
[25]张少文,张学成,王玲等.黄河上游年降雨径流预测研究[J].中国农村水利水电,2005,(1):41-45.
[26]任妮.改进L-M型BP模型的降雨径流预报[D].山东:大连理工大学,2006.
[27]任妮,金生.基于改进BP神经网络的降雨径流预报新方法[J].中国水运,2006,6(12):119-122.
[28]飞思科技产品研发中心编著. MATLAB6.5辅助神经网络分析与设计[M].北京:电子工业出版社,2003.
[29] MaierH.R., Dandy G.C. .Application of artificial neural networks to forecasting of surface water quality variables: issues, applications and challenges[M]. The Netherlands:Academic publishers,2003,1:284-309.
[30]研学论坛: http://bbs.matwav.com.
[31]蒋宗礼.人工神经网络导论[M].北京:高等教育出版社,2001.
[32]黑河网:http://www.yellowriver.gov.cn/trsweb/lyzl/lygh/200707/t20070720-24518.htm.
[33]施雅风,张祥松.气候变化对西北干旱区地表水资源的影响和未来趋势[J].中国科学(B辑),1995,25(9):968-977.
[34]蓝永超,丁永建,刘进琪.全球气候变暖情景下黑河山区流域水资源的变化[J].中国沙漠,2005,25(6):863-868.
[35]施雅风,沈永平,李栋梁,张果味等.中国西北气候由暖干向暖湿转型的特征和趋势探讨[J].第四季研究,2003,23(2):152-164.
[36]施雅风,沈永平,胡汝骥.西北气候由暖干向暖湿转型的信号——影响和前景初步探讨[J].冰川冻土,2002,24(3):219-226.
[37]李林,王振宇,汪青青.黑河上游地区气候变化对径流量的影响研究[J].地理科学,2006,26(1):40-46.
[38]丁永建,叶柏生,刘时银.祁连山区流域径流影响因子分析[J].地理学报,1999,54(5):431-437.
[39]胡铁松,袁鹏,丁晶.人工神经网络在水文水资源中的应用[J].水科学进展,1995,6(1):76-82.
[40]姚林声.径流长期预报的人工神经网络方法[J].水科学进展,1995,6(1):61-65.
[41]徐建华.现代地理学中的数学方法[M].高等教育出版社,2002.
[42]楚永伟,蓝永超,李向阳等.黑河莺落峡站年径流长期预报模型研究[J].中国沙漠,2005,25(6):869-873.
[43]唐启义,冯明光. DPS数据处理系统——实验设计.统计分析及模型优化[M].科学出版社,2006.
[44]刘少玉,卢耀如,程旭学.峡径流变化与人类活动的影响分析[J].地球学报,2005.
[45]肖生春,肖洪浪.年来人类活动对黑河流域水环境的影响[J].干旱区资源与环境,2004,18(3):57-62.
[46] Amell, Bates B, Lang H, etal. Hydrology and freshwater ecology[A]. In: Watson R. T., etal, eds, Climate change 1995[C]. Cambridge: Cambridge university Press, 1996,1:327-328.
[47]施雅风,刘时银.中国冰川对21世纪全球变暖影响的预估[J].中国科学(B辑),1995,25(9):968-977.
[48]柳景峰,张勃.西北干旱区近50年气候变化对出山径流的影响分析——以黑河流域为例[J].干旱区资源与环境,2007,21(8):58-63.
[49]康尔泗,程国栋,蓝永超等.西北干旱区内陆河流域出山径流变化趋势对气候变化响应模型[J].中国科学(D辑),1999,29(1):47-54.
[50]冯国章,李佩成.人工神经网络结构对径流预报精度的影响分析[J].自然资源学报,1998,13(2):169-174.
[51]李村军,邓红霞,朱兵等. BP神经网络预测日径流序列的数据适应性分析[J]. 2007,39(2):25-29.
[52]方红,郑文瑞.径流预报的模糊神经网络方法[J]. 2004,34:80-82.
[53]王文,马骏.若干水文预报方法综述[J].水利水电科技进展,2005,25(1):56-60.
[54]周娜,周燕萍.神经网络及遗传算法在径流预报中的应用[J].计算机仿真,2004,21(9): 117-119.
[55]马炼,王银堂,张闻胜.神经网络技术在水文系列中长期预报中的应用[J].水利水电技术,2002,33(2):5-7.
[56] R.D.辛格.用变化的降雨——径流资料通过反向传播人工神经网络模拟径流[J].水资源研究,2006,27(2):45-49.
[57]熊立华,郭生练.自适应神经模糊推理系统在水文模型综合中的应用[J].水文,2006,26(1):38-41.
[58]钟登华,王仁超,皮钧.水文预报时间序列神经网络模型[J].水利学报,1995(2):69-75.
[59]姚林声.径流长期预报的人工神经网络方法[J].水科学进展,1995,6(1):61-65.
[60] D.Hanselman, B.Littlefield著,张航,,黄攀译.精通.MATLAB6[M].清华大学出版社,2002.
[61]刘永,张立毅. BP和RBF神经网络的实现及其性能研究[J].电子测量技术,2007,30(4):77-80.
[62]刘良林,王全凤,林煌斌. BP神经网络参数设定及应用[J].基建优化,2007,28(2):101-106.
[63]刘向阳. BP神经网络建模与应用[J].计算机与网络,2007,16:163-165.
[64]王立柱,赵大宇. BP神经网络的改进及应用[J].沈阳师范大学学报(自然科学版),2007,25(1):61-64.
[65]胡宾,崔广柏,朱灵芝. BP神经网络预测河流月径流量[J].浙江水利科技,2007,(2):15-19.
[66]马光志,胡韶峰. BP算法稳定性与收敛性的一种改进[J].华中科技大学学报(自然科学版),2002,30(12):21-22.
[67]刘浩,白振兴. BP网络的MATLAB实现及应用研究[J].现代电子技术,2006,(2):49-51.
[68]董聪.前向网络全局最优化问题研究[J].中国科学基金,1:23-29.
[69]王鸿斌,张立毅.新型神经网络的发展及其应用[J].忻州师范院报,2007,23(2):50-53.
[70]刘艳荣.于预测的BP网络结构设计[J].农业网络信息,2007,5:208- 209.
[71]鲁子奕,杨绿溪,吴球等.提高前馈神经网络泛化能力的新算法[J].电路与系统学报,1997,2(4):7-12.
[72]王蕊,夏军.多支流河段径流预测方法比较研究[J].人民黄河,2007,29(3):29-32.
[73]李彩林,许慧君,王金文.数据挖掘技术在径流预报中的应用[J].水电自动化与大坝检测,2007,31(2):75-81.
[2] Pankaj Singh, M.C.Deo. Suitability of different neural networks in daily flow orecasting[J]. Applied Soft Computing, 2007(7):968-978.
[3] Cameron M., Zealand, Donald H., etal. Short term streamflow forecasting using artificial neural networks[J]. Journal of Hydrology, 1999(214):32-48.
[4] Sanaga, Srinivasulu, Ashu Jain. A comparative analysis of training methods for artificial neural network rainfall–runoff models[J]. Applied Soft Computing, 2006(6):295-306.
[5] S.Riad, J.Mania. Rainfall-Runoff model using an artificial neural network approach[J]. Mathematical and Computer Modelling, 2004(40):839-846.
[6] Mahmut Firat, Mahmud Giiongor. River flow estimation using adaptive neuro fuzzy inference system[J]. Mathematics and Computers in Simulation, 2006(14):32-48.
[7] Wen Wang, Pieter H.A.J.M., Van Gelder. Forecasting daily streamflow using hybrid ANN models[J]. Journal of Hydrology, 2006(324):383–399.
[8] Drasko Furundzic. Application example of neural networks four time series analysis[J]. Rainfall-runoff modeling, 1998(64):383-396.
[9] A.Castelletti, R.Soncini-Sessa. Bayesian Networks and participatory masodeling in water resource management[J]. Environmental Modelling & Software, 2007(22):1075-1088.
[10] Y.B.Dibike, D.P.Solomatlne. River flow forecasting using artificial neural networks[J]. Phys. Chem. Earth(B), 2001,26(1):1-7.
[11]张立明.人工神经网络的模型及其应用[M].复旦大学出版社,1993.
[12]王文,马骏.若干水文预报方法综述[J].水利水电科技进展,2005,25(1):56-60.
[13]吴险峰,刘昌明.流域水文模型研究的若干进展[J].地理科学进展,2002,21(4):341-348.
[14] Famili A., Shen Weimin. Data preprocessing and intelligent data analysis,intelligent data analysis[Z].
[15] Feng C., Miehie D. Machine learning of rules and trees in machine learning[M]. Neural and statistical Classification, 1994.
[16]王栋,曹升乐.人工神经网络在水文水资源水环境系统中的应用研究进展[J].水利水电技术,1999,30(12):4-7.
[17]杨荣富,丁晶,刘国东.神经网络模拟降雨径流过程[J].水利学报,1998(10):69-73.
[18]陈仁升,康尔泗,张济世.应用GRNN神经网络模型计算西北干旱区内陆河流域出山径流[J].水科学进展,2002,13(1):87-91.
[19]徐中民,蓝永超,程国栋.人工神经网络方法在径流预报中的应用[J].冰川冻土,2000,22(4):372-375.
[20]王玲,黄国如.基于径流分类的日径流预测神经网络模型[J].灌溉排水,2002,21(4):45-48.
[21]王玲,黄国如.改进的BP网络模型在流于日径流量预测中的应用[J].水电能源科学,2003,21(1):32-34.
[22]张婧婧,姜铁兵,康玲等.基于人工神经网络的日径流预测[J].水自动化和大坝监测,2002,26(4):65-67.
[23]尹雄锐.基于智能计算的降雨径流模拟方法研究[D].武汉大学,2004.
[24]张少文,张学成,王玲等.黄河年降雨-径流BP预测模型研究[J].人民黄河,2005,27(1):18-20.
[25]张少文,张学成,王玲等.黄河上游年降雨径流预测研究[J].中国农村水利水电,2005,(1):41-45.
[26]任妮.改进L-M型BP模型的降雨径流预报[D].山东:大连理工大学,2006.
[27]任妮,金生.基于改进BP神经网络的降雨径流预报新方法[J].中国水运,2006,6(12):119-122.
[28]飞思科技产品研发中心编著. MATLAB6.5辅助神经网络分析与设计[M].北京:电子工业出版社,2003.
[29] MaierH.R., Dandy G.C. .Application of artificial neural networks to forecasting of surface water quality variables: issues, applications and challenges[M]. The Netherlands:Academic publishers,2003,1:284-309.
[30]研学论坛: http://bbs.matwav.com.
[31]蒋宗礼.人工神经网络导论[M].北京:高等教育出版社,2001.
[32]黑河网:http://www.yellowriver.gov.cn/trsweb/lyzl/lygh/200707/t20070720-24518.htm.
[33]施雅风,张祥松.气候变化对西北干旱区地表水资源的影响和未来趋势[J].中国科学(B辑),1995,25(9):968-977.
[34]蓝永超,丁永建,刘进琪.全球气候变暖情景下黑河山区流域水资源的变化[J].中国沙漠,2005,25(6):863-868.
[35]施雅风,沈永平,李栋梁,张果味等.中国西北气候由暖干向暖湿转型的特征和趋势探讨[J].第四季研究,2003,23(2):152-164.
[36]施雅风,沈永平,胡汝骥.西北气候由暖干向暖湿转型的信号——影响和前景初步探讨[J].冰川冻土,2002,24(3):219-226.
[37]李林,王振宇,汪青青.黑河上游地区气候变化对径流量的影响研究[J].地理科学,2006,26(1):40-46.
[38]丁永建,叶柏生,刘时银.祁连山区流域径流影响因子分析[J].地理学报,1999,54(5):431-437.
[39]胡铁松,袁鹏,丁晶.人工神经网络在水文水资源中的应用[J].水科学进展,1995,6(1):76-82.
[40]姚林声.径流长期预报的人工神经网络方法[J].水科学进展,1995,6(1):61-65.
[41]徐建华.现代地理学中的数学方法[M].高等教育出版社,2002.
[42]楚永伟,蓝永超,李向阳等.黑河莺落峡站年径流长期预报模型研究[J].中国沙漠,2005,25(6):869-873.
[43]唐启义,冯明光. DPS数据处理系统——实验设计.统计分析及模型优化[M].科学出版社,2006.
[44]刘少玉,卢耀如,程旭学.峡径流变化与人类活动的影响分析[J].地球学报,2005.
[45]肖生春,肖洪浪.年来人类活动对黑河流域水环境的影响[J].干旱区资源与环境,2004,18(3):57-62.
[46] Amell, Bates B, Lang H, etal. Hydrology and freshwater ecology[A]. In: Watson R. T., etal, eds, Climate change 1995[C]. Cambridge: Cambridge university Press, 1996,1:327-328.
[47]施雅风,刘时银.中国冰川对21世纪全球变暖影响的预估[J].中国科学(B辑),1995,25(9):968-977.
[48]柳景峰,张勃.西北干旱区近50年气候变化对出山径流的影响分析——以黑河流域为例[J].干旱区资源与环境,2007,21(8):58-63.
[49]康尔泗,程国栋,蓝永超等.西北干旱区内陆河流域出山径流变化趋势对气候变化响应模型[J].中国科学(D辑),1999,29(1):47-54.
[50]冯国章,李佩成.人工神经网络结构对径流预报精度的影响分析[J].自然资源学报,1998,13(2):169-174.
[51]李村军,邓红霞,朱兵等. BP神经网络预测日径流序列的数据适应性分析[J]. 2007,39(2):25-29.
[52]方红,郑文瑞.径流预报的模糊神经网络方法[J]. 2004,34:80-82.
[53]王文,马骏.若干水文预报方法综述[J].水利水电科技进展,2005,25(1):56-60.
[54]周娜,周燕萍.神经网络及遗传算法在径流预报中的应用[J].计算机仿真,2004,21(9): 117-119.
[55]马炼,王银堂,张闻胜.神经网络技术在水文系列中长期预报中的应用[J].水利水电技术,2002,33(2):5-7.
[56] R.D.辛格.用变化的降雨——径流资料通过反向传播人工神经网络模拟径流[J].水资源研究,2006,27(2):45-49.
[57]熊立华,郭生练.自适应神经模糊推理系统在水文模型综合中的应用[J].水文,2006,26(1):38-41.
[58]钟登华,王仁超,皮钧.水文预报时间序列神经网络模型[J].水利学报,1995(2):69-75.
[59]姚林声.径流长期预报的人工神经网络方法[J].水科学进展,1995,6(1):61-65.
[60] D.Hanselman, B.Littlefield著,张航,,黄攀译.精通.MATLAB6[M].清华大学出版社,2002.
[61]刘永,张立毅. BP和RBF神经网络的实现及其性能研究[J].电子测量技术,2007,30(4):77-80.
[62]刘良林,王全凤,林煌斌. BP神经网络参数设定及应用[J].基建优化,2007,28(2):101-106.
[63]刘向阳. BP神经网络建模与应用[J].计算机与网络,2007,16:163-165.
[64]王立柱,赵大宇. BP神经网络的改进及应用[J].沈阳师范大学学报(自然科学版),2007,25(1):61-64.
[65]胡宾,崔广柏,朱灵芝. BP神经网络预测河流月径流量[J].浙江水利科技,2007,(2):15-19.
[66]马光志,胡韶峰. BP算法稳定性与收敛性的一种改进[J].华中科技大学学报(自然科学版),2002,30(12):21-22.
[67]刘浩,白振兴. BP网络的MATLAB实现及应用研究[J].现代电子技术,2006,(2):49-51.
[68]董聪.前向网络全局最优化问题研究[J].中国科学基金,1:23-29.
[69]王鸿斌,张立毅.新型神经网络的发展及其应用[J].忻州师范院报,2007,23(2):50-53.
[70]刘艳荣.于预测的BP网络结构设计[J].农业网络信息,2007,5:208- 209.
[71]鲁子奕,杨绿溪,吴球等.提高前馈神经网络泛化能力的新算法[J].电路与系统学报,1997,2(4):7-12.
[72]王蕊,夏军.多支流河段径流预测方法比较研究[J].人民黄河,2007,29(3):29-32.
[73]李彩林,许慧君,王金文.数据挖掘技术在径流预报中的应用[J].水电自动化与大坝检测,2007,31(2):75-81.