【作者】 曾剑；

【导师】 孙志林； 潘存鸿；

【作者基本信息】 浙江大学 ， 港口、海岸及近海工程， 2011， 博士

【摘要】 钱塘江河口潮强流急,涌潮汹涌,河床冲淤变化剧烈,是最具代表性的强潮河口。探索与运用钱塘江河口水沙运动规律,一直是治理和保护该河口的一项重要的基础性工作。历经几代人的努力,钱塘江河口水沙运动规律研究发展了实测资料分析、数值模拟计算与河工模型试验等多种预测手段并取得了系列的研究成果。近几年,灰色理论、神经网络、分形理论、小波分析等现代数据挖掘技术的出现,为水沙运动规律研究提供了新的研究思路与途径。论文在总结前人研究的基础上,利用钱塘江河口大量的实测资料,采用小波分析、人工神经网络及其藕合模型等数据挖掘技术,结合理论分析、水槽试验等研究手段,揭示钱塘江河口径流、涌潮、细颗粒泥沙起动流速与沉降速度、潮流挟沙能力以及河床冲淤等变化特征与运动规律,以便为钱塘江河口的整治及重大涉水工程建设提供技术基础,主要工作内容归纳如下：(1)基于人工神经网络的预测模型研究。针对钱塘江涌潮高度、泥沙起动流速、细颗粒泥沙絮凝沉降速度及潮流挟沙能力与其影响因素之间存在复杂的非线性关系,在识别各影响因子与分析训练样本代表性的基础上,运用神经网络的最佳函数逼近性能,建立了基于径向基函数(RBF)网络的预测模型。经实测与试验数据的检验,该类模型均具有较好的泛化能力,并在定量评价各影响因子的贡献程度中得到了成功应用。(2)基于小波分析的水动力与河床冲淤的长周期特性研究。依据长系列的水文与地形资料,在分析复Morlet小波函数基本特性的基础上,选用该小波对标准化序列进行了小波变换,根据小波变换系数与小波方差,揭示了钱塘江河口年径流、潮差、沙坎高程以及河床容积的变化过程伴随着21a左右的显著周期,预计2010年以后,该河口径流将转入偏丰期。并以显著周期为纽带,提出了钱塘江河口水沙运动的长周期演变模式。(3)涌潮高度对影响因素与治江缩窄工程的响应研究。通过涌潮高度的神经网络模型,结合正交试验设计理论,定量地分析了各因素与大规模治江缩窄工程对涌潮的影响程度,结果表明,涌潮高度与下游潮汐呈正相关关系,而与河道地形呈负相关关系,当径流流量小于7000 m3/s时,涌潮高度随径流增大而增大,超过此范围后,则随径流增大而减小。大规模治江缩窄后,涌潮沿程强度分布趋于稳定,在顺直江道条件下,盐官涌潮高度减小了0.17m,而弯曲江道,则增大了0.39m。(4)细颗粒泥沙起动流速、沉降速度与潮流挟沙能力的经验公式研究。基于大量的水沙实测资料与试验数据,通过量纲分析与多元回归率定相关参数,建立了适合于钱塘江河口水流特性的细颗粒泥沙起动流速、沉降速度与潮流挟沙能力的经验拟合公式。经实测与试验资料检验,公式具有良好的拟合精度与应用范围。此外,文中还提出了从非恒定实测水沙过程中提取泥沙起动流速与潮流挟沙能力的方法。(5)基于小波分析与神经网络的河床容积预报模型研究。针对河床容积变化具有趋势性、周期性和随机性的特点,在小波分析将时间序列分解成若干细节序列和背景序列的基础上,结合人工神经网络技术,建立了钱塘江河口河床容积的预报模型。经实测资料检验,最优的预报模型(方案6),其拟合阶段与检验阶段的合格率可达97.5%与88.2%,确定性系数可达0.90与0.71,计算结果总体令人满意,为河床容积预报提供了一种可行的技术方案。更多还原

【Abstract】 The Qiantang River estuary is the most representative estuary with strong tide, because of it’s strong tide, tidal bore and the tempestuously change of erosion-deposit on the river bed. It is always an important basic work of estuary treatments and protection that researching and using the hydraulic characteristics of water and sediment movement at the Qiantang River estuary. The hydraulic characteristics aforementioned are developed by methods, such as analysis of field data, numerical model and physical model and a series of achievements are obtained basing on the past studies. Recently, the modern data analysis technologies, such as gray theory, neural network, fractal theory and wavelet analysis, provide a new approaching methods and thinking for the studies on hydraulic characteristics.In this paper, the characteristics of run off of the Qiantang estuary, bore, sediment-moving incipient velocity, sedimentation velocity of fine sediment, the sediment carrying capacity of tide and riverbed scouring-siltation are discussed by new data analysis technologies, wavelet analysis, neural network and the combination of the later two methods, theoretical analysis and experimental research, basing on the past studies. And it also provides a technological support for the Qiantang River estuary treatment and important wading water management projects. The mainly work contents are as below:(1) The model for predicting the height of tidal bore basing on neural network. According to complex non-linear relationship between the tidal height of the Qiantang bore, sediment-moving incipient velocity, flocculation and sedimentation velocity of fine sediment and sediment carrying capacity of tide and their influence factors, a predicting model is established by RBF (Radial Basis Function) basing on identifying the major influence factors and improving the representativeness of learning sample. The comparison of the simulated results with the observation data shows that the model has good generalization ability and it is successfully adopted to quantitatively analyze the contributions of each influence factor.(2)Study on hydrodynamic force and long-periodic feature of riverbed scouring-siltation by wavelet analysis. In accordance with the non-stationary hydrological time series and topographical data, the wavelet method was adopted to analyze the periodic feature and the trend of the runoff in the Qiantang estuary. The complex Morlet function, whose basic characteristic is discussed, is selected to transform the standardized time-series of the annual runoff discharge. Basing on the distribution of module and real of wavelet transform coefficients and wavelet variance, the prominent period of the annual runoff discharge in the Qiantang River estuary is 21 years approximately, and it is predicted that the trend of the annual flow is in the low flow period till 2010, and then the annual discharge is increase. According to the prominent period, the long-term evolution model of water and sediment movement at the Qiantang River estuary is proposed.(3)Study on response of height of tidal bore to influence factors and regulation works. Based on qualitatively analyzing influence factors, the model for predicting the height of tidal bore at Yanguan is established by using BP neural network. The comparison of the simulated results with the observation data shows that the model has good generalization ability. Combining the theory of orthogonal experiment design, the major affecting factors such as tide, runoff and river topography are quantitatively analyzed. It is found that the height of tidal bore is positively related to the tide, and negatively related to the river topography. The height of tidal bore is positively related to the runoff when the discharge is less than 7000m3/s. However they are negative correlation relationship if the discharge is greater than 7000m3/s. Then the model is used in a case to study on the effect of large-scale reclamation and regulation works on the tidal bore in the Qiantang estuary. The influence extent to the bore height at Yanguan is quantitatively analyzed under different hydrodynamic and river-bed condition. The results show that the bore height decreased by 0.17m after reclamation under straight alignment of river course, and increased by 0.39m under curved alignment of river course.(4) Study on empirical formula of incipient velocity of fine sediment, sedimentation velocity, sediment carrying capacity of tide. Basing on Large amounts of field data and experimental results, the empirical formulas are established by dimensional analysis and multiple regression analysis, which adapt the flow characteristics of Qiantang River estuary. Comparison with the field data and experimental results, it indicates that computation results by the formulas have a good agreement and application extent. In this paper, a method for deciding incipient velocity of sediment and sediment carrying capacity of tide through a non-stationary flow and sediment process is established.(5)Study on prediction model for riverbed volume by wavelet analysis and neural network. Due to the change of riverbed volume has characteristics of tendency, periodicity, randomness, a prediction model for riverbed volume of the Qiantang River estuary is established by using the technologies of wavelet analysis and neural network. Comparison with the field data, the model reaches 97.5% and 88.2% accuracy rate in fitting and testing process and 0.90 and 0.71 coefficient of determination, respectively. The computation results are acceptable and it provides a feasible technical solution for riverbed volume prediction.更多还原