南洞地下河月径流时间序列的混沌特征及预测

利用基于相空间重构技术、混沌识别与预测理论对1993-2013年南洞地下河月径流时间序列的非线性特征进行了分析,由所获得的延迟时间和最佳嵌入维数实现了月径流时间序列的相空间重构,运用饱和关联维数法和小数据量法计算出南洞地下河月径流时间序列的饱和关联维数和最大Lyapunov指数,并运用Volterra模型对南洞地下河月径流时间序列进行了多步预测研究。研究结果表明,南洞地下河月径流时间序列相空间重构的延迟时间和最佳嵌入维数分别为τ=5、m=8,饱和关联维数D和最大Lyapunov指数λ分别为4.63、0.748 9,从定性和定量的角度证明了南洞地下河月径流时间序列具有弱混沌特征。Volterra自适应滤波模型的预测结果能较好地表征南洞地下河月径流的变化趋势和规律,对18个月内的短期预测精度较高,模拟效果较好。      

混沌时间序列支持向量机模型及其在径流预测中应用

以重构相空间理论为基础,探讨了混沌时间序列的支持向量机预测模型建模的思路、特点及关键参数的选取。利用饱和关联维数法进行相空间重构,并运用改进小数据量法计算最大Lyapunov指数,对宜昌站月径流时间序列进行混沌特性识别。在运用混沌时间序列的支持向量机模型对月径流预测的应用中,引入了径向基核函数,简化了非线性问题的求解过程。实例表明,该模型能较好地处理复杂的水文序列,具有较高的泛化能力和很好的预测精度。     

0-1测试方法的径流时间序列混沌特性应用

径流时间序列混沌特性识别的常用方法是基于相空间重构的关联维数法、最大Lyapunov指数法和Kolmogorov熵法。引入一种新的时间序列混沌特性识别方法:0-1混沌测试方法。该方法直接应用于时间序列不需要相空间重构,并且通过量化指标K_c是否接近于0或1来识别时间序列的混沌特性。以Logistic映射生成的序列、金沙江流域和美国Umpqua河多年日径流序列为研究对象,首先利用0-1混沌测试方法进行了混沌特性识别和判定;然后基于相空间重构,运用相空间重构、伪最近邻点法、关联维数方法、最大Lyapunov指数法和Kolmogorov熵5种非线性研究方法分析了这两列径流时间序列混沌特性。研究结果表明0-1混沌测试方法简单有效。以上方法交互验证了该两列径流时间序列存在低维混沌特性。     

黑河日径流量混沌变化特性的研究—Ⅱ相空间嵌入维数的确定

通过对黑河日径流量变化规律的探索,重点论述相空间嵌入维数的确定方法。计算结果表明：黑河日径流量时间序列的关联维数为2.1,是低维分数;另外,采用饱和关联积分法,得到黑河日径流量序列的嵌入维数为9。     

嫩江流域径流的可预报性及其时间尺度分析

本文在阐述径流过程主要影响因素和表现特征基础上,采用混沌理论,对于嫩江流域月径流过程系统进行了延滞时间、嵌入纬度、关联维数等计算,进而进行相空间重构,计算出其最大Lyapunov指数为0.13,说明该系统具有混沌特征.初步得出如下结论:①嫩江流域中长期径流预报的预见期为7～8个月;②"数据驱动模型"对于中长期径流预报更有其方法的适用性;③中长期径流预报计算时段应以月为单位.     

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

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。     

库岸滑坡地下水位时间序列混沌特征识别与PSO-LSSVM模型预测

地下水位预测对滑坡稳定性分析具有重要意义,三峡库区库岸滑坡地下水位时间序列在季节性强降雨和周期性库水位涨落等诸多因素影响下呈现混沌特征。在对地下水位序列进行相空间重构的基础上,采用饱和关联维数法和最大Lyapunov指数法对其混沌特征进行验证。再用预测性能优秀的最小二乘支持向量机(LSSVM)模型对其进行预测,并用粒子群算法优化选取LSSVM模型的参数,以克服LSSVM模型参数选取困难的缺点。以三峡库区三舟溪滑坡前缘STK-1水文孔日平均地下水位序列为例进行了混沌分析,分别运用粒子群优化的LSSVM模型(PSO-LSSVM)和BP神经网络模型对STK-1水文孔地下水位进行了预测。结果表明库岸滑坡地下水位序列存在混沌特征,PSO-LSSVM模型预测结果的均方根误差为0.193m,拟合优度为0.815,说明预测效果较理想,且PSO-LSSVM模型预测精度高于BP网络模型,具有较强的实用性。     

基于GA-SVM的矿井涌水量预测

矿井涌水量的准确预测对预防矿山透水事故的发生至关重要,提出利用GA优化的SVM模型（GA-SVM）来实现矿井涌水量的短期准确预测。该方法利用GA的自动寻优功能寻找SVM的最佳参数,提高了预测的准确率。首先,利用微熵率法求矿井涌水量时间序列的最佳嵌入维数和延迟时间,进行相空间重构。其次,采集义煤集团千秋煤矿2011—2015年实际涌水量的时间序列,利用GA-SVM模型对最后12组数据进行预测,其预测平均绝对百分比误差仅为0.92%,最大相对误差为2.62%。最后,与PSO-SVM和BP神经网络预测进行对比,结果表明GA-SVM优化模型适用于矿井涌水量的预测并且预测精度较高。      

2.5Ma 以来古气候变化的混沌吸引子

利用2.5Ma以来经过时标调谐的宝鸡黄土剖面的粒度比值曲线,等时距(1000a)地重构了具2500个样本容量的时间序列。作者在分析此时间序列所记录的古气候变化的混沌特征后发现:2.5Ma以来的古气候是具有有限个自由度的复杂的混沌系统,吸引子关联维数为3.8,饱和嵌入维数为11,它的以最大Lyapunov指数和二阶Renyi熵表征的可预报时效(误差增大一倍所需时间)分别为0.66—0.76Ma和6600—7100a。已有的研究证明该剖面古气候时间序列的大部分时间区间受地球轨道参数的驱动,表明2.5Ma以来的古气候是一种混沌的与周期性的混合式振荡的动力系统。     

边坡演化的非线性时间序列多元混沌判别

以实测非线性时间序列为对象, 通过估计延迟时间与嵌入维数的相空间重构方法, 采取排除时间相关性点对的方法计算边坡系统关联维数D2;采用改进的Kantz法计算最大Lyapunov指数、以K₂熵作为Kolmogorov熵的近似, 并引入近似熵ApEn及系统复杂度混沌特征指标, 研究了边坡演化的多元混沌特征.通过实例分析, 发现多数边坡系统关联维数D₂为非整数, 最大Lyapunov指数、熵值均大于零以及系统复杂度位于(0, 1) 区间偏小值, 通过与确定性系统特征量的比较, 揭示了边坡系统的混沌特征, 并得出临滑阶段边坡混沌特征最为明显的结论.      

人工神经网络方法在径流预报中的应用

采用ＢＰ神经网络模型,以西北内陆河黑河流莺峡年平均出山地表流量为研究对象,对人工神经网络研究方法在干旱区环流径流预报中的应用进行了初步尝试,结果表明该预报成功率较高,证实了人工神经网络方法应用于流量预报领域的可行性,并分析了该方法在预报过程中的优缺性。     

径流长期预报的人工神经网络方法

提出径流长期预报的人工神经网络方法。运用神经网络的一典型模型──“反向传播”模型,以大伙房水库在补水期的径流状况作为研究对象,尝试了神经网络方法的效果。结果表明,该方法预报成功率较高,容错能力较强,可望成为径流长期预报的有效的辅助手段。     

The application of artificial neural networks to simulate meltwater runoff of Keqikaer Glacier,south slope of Mt. Tuomuer,western China

Because glacial melting provides a significant amount of surface water resources, especially in cold arid regions, it is critical that effective methods be developed for predicting their behavior. Glacier runoff differs from other types of stream flows, being characterized by large diurnal fluctuations, with maximum discharge during the summer months. Moreover, the size and remoteness of glaciers makes them difficult to study directly. Hence, developing effective modeling techniques is our best hope for understanding and predicting glacial melting phenomena. In the past, physics-based models have been used with some success. In this study, conducted in 2003 and 2004 on the Keqikaer Glacier on the south slope of Mt. Tuomuer, however, we used the newer artificial neural networks (ANNs) modeling technique. As the input nerve cell, we used the hourly wind speed, precipitation, air temperature, radiation balance, and ground temperature; the output nerve cell was the diurnal runoff at the glacial terminus. We then analyzed the simulated results under different scenarios by varying the input-nerve-cell parameters. It was found that ANN can simulate the process of glacier meltwater runoff successfully when basic parameters such as air temperature, precipitation and radiation balance are few. The results indicate that ANN can simulate the process of glacial meltwater runoff quite well, and that meteorological variables could in fact be used successfully to simulate glacier meltwater runoff using the ANN method.     

黑河日径流量混沌变化特性研究

通过对黑河日径流量变化规律的探索,重点论述相空间嵌入滞时的确定方法。计算结果表明：黑河日径流量的自相关函数呈现出明显的周期性,其周期与日径流量数据序列的周期基本一致;另外,采用了曲线下降趋势变缓的转折点,得到黑河日径流量序列的时间延迟为10天。     

Prediction of thermal conductivity of rocks by soft computing

The transfer of energy between two adjacent parts of rock mainly depends on its thermal conductivity. Knowledge of the thermal conductivity of rocks is necessary for the calculation of heat flow or for the longtime modeling of geothermal resources. In recent years, considerable effort has been made to develop artificial intelligence techniques to determine these properties. Present study supports the application of artificial neural network (ANN) in the study of thermal conductivity along with other intrinsic properties of rock due to its increasing importance in many areas of rock engineering, agronomy, and geoenvironmental engineering field. In this paper, an attempt has been made to predict the thermal conductivity (TC) of rocks by incorporating uniaxial compressive strength, density, porosity, and P-wave velocity using artificial neural network (ANN) technique. A three-layer feed forward back propagation neural network with 4-7-1 architecture was trained and tested using 107 experimental data sets of various rocks. Twenty new data sets were used for the validation and comparison of the TC by ANN. Multivariate regression analysis (MVRA) has also been done with same data sets of ANN. ANN and MVRA results were compared based on coefficient of determination (CoD) and mean absolute error (MAE) between experimental and predicted values of TC. It was found that CoD between measured and predicted values of TC by ANN and MVRA were 0.984 and 0.914, respectively, whereas MAE was 0.0894 and 0.2085 for ANN and MVRA, respectively.     

黑河日径流量混沌变化特性的研究——Ⅲ最大李雅普诺夫(Lyapunov)指数的确定

通过对黑河日径流量变化规律的探索,重点论述最大李雅普诺夫指数的确定方法,计算结果表明:黑河日径流量时间序列的最大李雅普诺夫指数为0.002012,大于零,可以判定黑河日径流量时间序列具有混沌特性.     

An artificial neural network model for flood simulation using GIS: Johor River Basin, Malaysia

Flooding is one of the most destructive natural hazards that cause damage to both life and property every year, and therefore the development of flood model to determine inundation area in watersheds is important for decision makers. In recent years, data mining approaches such as artificial neural network (ANN) techniques are being increasingly used for flood modeling. Previously, this ANN method was frequently used for hydrological and flood modeling by taking rainfall as input and runoff data as output, usually without taking into consideration of other flood causative factors. The specific objective of this study is to develop a flood model using various flood causative factors using ANN techniques and geographic information system (GIS) to modeling and simulate flood-prone areas in the southern part of Peninsular Malaysia. The ANN model for this study was developed in MATLAB using seven flood causative factors. Relevant thematic layers (including rainfall, slope, elevation, flow accumulation, soil, land use, and geology) are generated using GIS, remote sensing data, and field surveys. In the context of objective weight assignments, the ANN is used to directly produce water levels and then the flood map is constructed in GIS. To measure the performance of the model, four criteria performances, including a coefficient of determination (R ²), the sum squared error, the mean square error, and the root mean square error are used. The verification results showed satisfactory agreement between the predicted and the real hydrological records. The results of this study could be used to help local and national government plan for the future and develop appropriate (to the local environmental conditions) new infrastructure to protect the lives and property of the people of Johor.     

Performance evaluation of artificial neural networks for planimetric coordinate transformation—a case study,Ghana

Two national horizontal geodetic datums, namely, the Accra and Leigon datum, have been the only available datum used in Ghana. These two datums are non-geocentric and were established based on astro-geodetic observations. Relating these different geodetic datums mostly involves the use of conformal transformation techniques which could produce results that are not very often satisfactory for certain geodetic, surveying and mapping purposes. This has been ascribed to the incapability of the conformal models to absorb more of the heterogeneous and local character of deformations existing within the local geodetic networks. Presently, application of new approaches such as artificial neural network (ANN) is highly recommended. Whereas the ANN has been gaining much popularity to solving coordinate transformation-related problems in recent times, the existing researches carried out in Ghana have shown that only three-dimensional conformal transformation methods have been utilized. To the best of our knowledge, plane coordinate transformation between the two local geodetic datums in Ghana has not been investigated. In this paper, an attempt has been made to explore the plane coordinate transformation performance of two different ANN approaches (backpropagation neural network (BPNN) and radial basis function neural network (RBFNN)) compared with two different traditional techniques (six- and four-parameter models) in the Ghana national geodetic reference network. The results revealed that transforming plane coordinates from Leigon to Accra datum, the RBFNN was better than the BPNN and traditional techniques. Transforming from Accra to Leigon datum, both the BPNN and RBFNN produced closely related results and were better than the traditional methods. Therefore, this study will create the opportunity for Ghana to recognize the significance and strength of the ANN technology in solving coordinate transformation problems.