Knowledge Resource Center for Ecological Environment in Arid Area
基于数据挖掘的黑河中游地下水位动态分析及观测网优化设计 | |
其他题名 | Dynamic Variation Analysis of Groundwater and the Optimization of Groundwater Monitoring Network Based on Data Mining in the middle reaches of Heihe River |
于海姣 | |
出版年 | 2016 |
学位类型 | 硕士 |
导师 | 王涛 |
学位授予单位 | 中国科学院大学 |
中文摘要 | 在干旱区,地下水是重要的水资源,对区域的生态平衡具有重要的作用。掌握干旱区地下水位的动态特征,对地下水位进行准确的预测,是干旱区地下水资源管理和保护的重要基础;地下水动态观测网提供的地下水动态信息是掌握地下水系统变化的主要手段,也是研究地下水动态特征的基础数据来源。但由于技术水平等因素的限制,干旱区地下水观测井网络存在大量冗余性数据,如何利用一定的优化方法在不影响原有观测井网提供有效信息能力的基础上消除冗余观测井是干旱区地下水观测井网优化的重要内容。本文以黑河中游地区为研究区域,在收集和整理研究区水文、气象数据以及地下水监测数据的基础上,基于自组织特征映射网络、小波分析、人工神经网络、支持向量机等数据挖掘技术,对黑河中游地下水动态进行了分析并对地下水观测网进行了优化。主要工作内容归纳如下:(1)黑河中游地下水动态分类分析。以黑河中游地下水动态变化为基础,运用自组织特征映射网络(SOM)对黑河中游地下水位动态进行了分类并分析了不同类型的地下水位动态变化特征。根据黑河中游地区的水文地质条件,应用SOM方法可以将黑河中游地下水位动态分为开采型、径流补给型、灌溉影响型及蒸发型4类。分类结果表明,SOM方法能够模拟人脑自组织、自学习的特点,分类方法较为客观。该方法的应用可以为黑河中游地下水位动态分类提供新的思路与方法。(2)黑河中游地下水位动态与影响要素多尺度交叉小波分析。运用连续小波变换及交叉小波分析,探讨了黑河中游地下水位动态变化与区域气温、降水、蒸发及径流量等气象、水文因子在时间域和频率域中包含的周期性特征和多尺度相关关系。分析表明,气温、降水、蒸发、径流量和地下水位之间存在一个8~16a的共振周期;气温、降水、蒸发对地下水位动态的影响存在时间超前效应;径流量对地下水位的变化表现为显著的负相关关系,地下水位对径流量较气温、降水、蒸发更敏感,地表径流量是影响地下水位变化的直接因素;气温、降水、蒸发和径流量对地下水位的影响约在2000-2003年前后发生突变,这可能与黑河中游实施的生态输水计划有关;气温、降水、蒸发、径流量与地下水位动态的趋势和突变呈基本一致的变化特征。研究表明连续小波变换、交叉小波功率谱、交叉小波凝聚谱及位相关系对研究地下水动态的影响机制具有较强的实用价值,可以作为地下水动态影响机制研究的有效工具。(3)基于小波变换与人工神经网络耦合(WA-ANN)的黑河中游地下水位动态模拟预测。针对地下水位动态与气温、降水、蒸发、径流量之间复杂的非线性关系,将小波分析与人工神经网络相耦合建立WA-ANN模型,对观测井未来1个月的地下水位进行预测,将预测结果与未经小波变换的ANN模型进行对比以检验WA-ANN模型的有效性。结果显示,ANN和WA-ANN均可用于干旱区地下水位短期预测,但WA-ANN模型有更高的预测精度,WA-ANN模型在黑河中游地下水位预测方面具有更好的适用性;ANN和WA-ANN模型在极值预测部分误差较大,都不能很好的模拟地下水位极值。研究表明用小波变换对非平稳地下水位动态进行预处理具有一定的优越性,小波分析与人工神经网络相耦合共同反映出了地下水位的未来变化趋势。该方法的应用可以为干旱地区地下水位动态预测分析提供新的方法和思路,在资料有限的条件下成为地下水位预测的有效方法。(4)黑河中游地下水动态观测网优化。在分析研究区水文地质的条件下,通过支持向量机(SVM)网络建立黑河中游地下水观测网优化模型来实现黑河中游地下水观测井网的优化,去除冗余观测井。模型建立后,分别从优化模型SVM的评价指标、优化后地下水观测井的分布及优化后的地下水位流场三个方面对比分析了不同优化参数Epsilon影响下的黑河中游地下水观测网优化结果。结果表明,SVM模型的参数Epsilon控制着优化网络的大小与优化结果的精度,当SVM模型的优化参数Epsilon为0.01,即优化后地下水观测井为26个时,对应的SVM模型的模拟精度最高,模拟值和实测值之间误差最小;优化后的地下水观测井在在空间上均匀分布于黑河中游地区,能够满足区域地下水观测的需求;利用这26个观测井进行空间插值的地下水流场形态与优化前的地下水流场形态分布非常相似,无实质性的差别。研究结果表明SVM方法能够用于干旱区地下水观测井网的优化设计,可以为地下水观测网优化提供新的思路。 |
英文摘要 | Groundwater, which plays an important role in regional ecological balance, is the most important water resource and also the basis for constitution, development and stability of oasis ecosystem in arid regions. The dynamic characters of the groundwater level and the accurate prediction of the groundwater level are important basis for groundwater resources management. Groundwater dynamic information provided by the dynamic observation network is the main way to grasp the change of the groundwater system and is also the basic data source for the study of the dynamic characters of groundwater. However, due to technical limitations and other factors, groundwater observation network exits a lot of redundant data. Therefore, using certain optimizations to eliminate redundant observation wells without affecting the capacity to provide effective information of the original observation wells is one of the important goals of the groundwater observation network optimization.In this paper, the dynamic characters of groundwater level was analyzed and the groundwater observation network was optimized in the middle reaches of Heihe River using data mining technology, like self-organizing feature map neural network, wavelet analysis, artificial neural network and support vector machine. Data in this study included the groundwater observation data, hydrological and meteorological data of the middle reaches of Heihe River. The main study contents and conclusions are as followings:(1) Dynamic classification analysis of the groundwater level in the middle reaches of Heihe River. Based on the dynamic change of groundwater in the middle reaches of Heihe River, the groundwater level was classified by using self-organizing feature map (SOM) and the dynamic characters of different types of groundwater level were analyzed. According to the hydrogeological conditions of the middle reaches of Heihe River, the groundwater level in the middle reaches of Heihe River can be divided into 4 types, such as exploitation type, runoff-supply type, irrigation-influence type and evaporation type with the method of SOM. Classification results show that the SOM method can simulate human brain with the features of self-organized, self-learned and self-classified. SOM can provide new idea for the dynamic classification of groundwater level in the middle reaches of Heihe River.(2) Multi-scale cross wavelet analysis of groundwater level dynamics and influencing factors in the middle reaches of Heihe River. The relationships between the dynamic change of groundwater level and regional mean annual temperature, precipitation, evaporation and runoff in the middle reaches of Heihe river were analyzed using the method of continuous wavelet transform, cross wavelet analysis; multi time scale correlation in time and frequency domain and the periodic characters of dynamic change between groundwater level and influencing factors were discussed using wavelet cross correlation coefficient, wavelet coherence spectrum and the phase difference. Results showed that there is a resonant periodic of 8~16a in temperature, precipitation, evaporation, runoff and groundwater level; impacts of temperature, precipitation and evaporation on groundwater level have displayed time lag to some extent; the impacts of runoff on groundwater level showed a significant negative correlation, the groundwater level change is more sensitive to the runoff than temperature, precipitation and evaporation, surface runoff is the direct factor that influences the change of groundwater level; Effect of temperature, precipitation, evaporation and runoff on the groundwater level sudden changes in year 2000-2003, which may be influenced by the ecological water transport plan applied in the middle reaches of Heihe River. Studies have shown that continuous wavelet transform, wavelet cross power spectrum, cross wavelet coherence spectrum and phase relationship have strong practical values for the study of groundwater dynamic and can be used as effective tools to the study of groundwater dynamic mechanism.(3) Prediction of groundwater depth in the middle reaches of Heihe River based on wavelet-artificial neural network (WA-ANN). Prediction model of groundwater depth predicted monthly groundwater depth in 4 typical groundwater monitoring wells was constructed using WA-ANN to predict groundwater depth for one month ahead. In order to test the validity of the developed model, detailed comparisons were made between the WA-ANN model and the ANN model in terms of different evaluation criteria. Results showed that performances obtained by both the ANN and WA-ANN were satisfactory and WA-ANN model performed better than ANN model. The error of extreme values prediction in ANN and WA-ANN model is relatively large. Finally, it can be concluded that the WA-ANN model we had developed may be considered as an effective tool to establish a short-term monthly groundwater depth forecasting model in arid regions where have few meteorological observatories.(4) Optimization of groundwater observation network in the middle reaches of Heihe River. Based on the analysis of the hydrogeological conditions, the groundwater observation wells optimization model is established by support vector machine (SVM) model to remove redundant wells. After the model was established, the optimization results were analyzed from three aspects: the evaluation index of the optimization model SVM, the distributions of groundwater observation well and the groundwater flow field. Results showed that the parameter of SVM model, Epsilon, controls the size of the optimized network and the accuracy of the results. When Epsilon is 0.01, which means 26 optimized groundwater observation wells, the accuracy of the SVM model is highest and the errors between simulated value and measured value are smallest; the optimized groundwater observation wells are distributed evenly in the middle reaches of the Heihe River; the shape of the groundwater flow field is very similar to that before optimization. Therefore, it is sure that the optimization results can meet the requirements of dynamic analysis of groundwater level in the middle reaches of Heihe River. The research results show that the SVM method can be applied to the optimization of the groundwater network in arid regions. The SVM model provides a new method for the optimization of groundwater observation network. |
中文关键词 | 黑河中游 ; 数据挖掘 ; 地下水动态 ; 地下水观测网优化 |
英文关键词 | Middle reaches of Heihe River,groundwater dynamic, groundwater observation network optimization, data mining |
语种 | 中文 |
国家 | 中国 |
来源学科分类 | 自然地理学 |
来源机构 | 中国科学院西北生态环境资源研究院 |
资源类型 | 学位论文 |
条目标识符 | http://119.78.100.177/qdio/handle/2XILL650/287713 |
推荐引用方式 GB/T 7714 | 于海姣. 基于数据挖掘的黑河中游地下水位动态分析及观测网优化设计[D]. 中国科学院大学,2016. |
条目包含的文件 | 条目无相关文件。 |
个性服务 |
推荐该条目 |
保存到收藏夹 |
导出为Endnote文件 |
谷歌学术 |
谷歌学术中相似的文章 |
[于海姣]的文章 |
百度学术 |
百度学术中相似的文章 |
[于海姣]的文章 |
必应学术 |
必应学术中相似的文章 |
[于海姣]的文章 |
相关权益政策 |
暂无数据 |
收藏/分享 |
除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。