激流怪潮发生机理研究与预警统计模式研究

发布时间：2018-03-14 13:21

本文选题：怪潮激流　切入点：地形因素　出处：《上海交通大学》2013年硕士论文　论文类型：学位论文

【摘要】：苏北浅滩南通海域附近在最近几年经常会发生海难事故，事故过程类似。海水经常莫名其妙地汹涌而至，而且涨潮速度很快。这给生活在该区域的渔民的生命和财产安全带来了极大的隐患。而这种看似诡异的涨潮也被民间百姓称为“怪潮”。为了解决“怪潮”给当地渔民所带来的恐惧问题，国家海洋局组织了大量的人力、物力以及财力，重点突击解决这一高难度问题。本课题组作为该国家海洋公益性项目的子课题承担单位之一，提出了采用数理统计的数学方法针对该区域发生的事故原因进行深入分析。主要的研究成果如下所示。根据该浅滩区域的高分辨率和低分辨率的遥感影像图，可以清晰地看出该近海区域地形地貌的变化。主要是由于河流的搬运作用，将陆源沉积物不断向海岸搬迁。根据高分辨率的遥感影像图，可以发现，人为的近岸滩涂围垦也向海域延伸，这也就加剧了该区域地形的变动。针对地形对“怪潮”发生的影响，建立了沟槽流速通量分析模型。通过ArcGIS软件对地形数据进行了插值分析，可以发现，苏北浅滩的地形特殊，成多条沙脊状，到近海地带，地形会变浅变窄。根据流量与流速的量化关系，采用MATLAB、EXCEL等数据编程软件，量化得出了近海岸流速会增加的结论。在使用统计学方法对海洋要素数据进行分析，主要分成以下三种模型：多变量自回归模型；人工神经网络模型；非线性混沌动力学模型。多变量自回归模型（AR模型），是将各个按时间尺度排列的海洋要素，按照时间序列的处理方法，寻找到最适的模型阶数。这也就得出能够影响当前的海洋要素数值的时间尺度。根据最适模型阶数，采用极大似然法计算出模型的自相关系数，建立起AR模型。模型的自相关系数矩阵，，就充分反映出各个海洋要素在不同时间维度上的相互之间的线性定量关系。通过统计的方法寻找到流速突变的原因。在本文重点的研究时间段内，风速会对潮位产生一定的影响，会造成潮位增加或者减小，而潮位的变化会对海洋的流速产生影响。人工神经网络模型的目的是，通过建立时间延迟网络，寻找出各个海洋要素在不同时间维度上相互之间的非线性关系。神经网络的特点是，在每个神经元内所发生的计算都很简单，但是每个简单的神经元一层接一层地连接起来，就形成了庞大复杂的神经网络，犹如脑神经一般。在进行数据拟合的过程中，多变量的输入向量使用神经网络进行拟合，效果并不好，不如单变量的输入向量拟合效果好。而采用神经网络模型进行预测，效果也并不理想，这主要是因为网络中权重个数太多，使得预测出的结果会发生很大的偏差。非线性混沌动力学模型，也是寻找出各个海洋要素之间的非线性关系。还有一个重要作用就是计算出分析数据的Lyapunov指数和kolmogrov熵。根据这两个指标可以得出预测的可信时间尺度。在本研究中，东西方向和南北方向分别进行讨论。东西向的风速、潮位以及流速拟合以及预测效果都比较好，而且预测的可信时间尺度也较大；而南北向的风速、潮位以及流速的可信时间尺度较短，更易出现混沌现象。
[Abstract]:Near the northern Jiangsu shoal in Nantong waters in recent years often happen accidents, the accident process is similar. The sea land often rather baffling raging, and the speed is very fast. This has brought great hidden danger to life in the region of the fishermen's life and property safety. This seems strange tide is also called people "the strange tide". In order to solve the problem of fear the strange tide brought to the local fishermen, the State Oceanic Administration Organization a lot of manpower, material resources and financial resources, the key to solve the difficult problem of assault.
As one of the sub subjects of the National Marine public welfare project, our research group put forward mathematical statistics method to analyze the causes of the accident in the area. The main research results are as follows.
According to the shallow areas of high and low resolution remote sensing images, we can clearly see that the coastal regional topography changes. Mainly due to handling of the river, the terrigenous sediments to coast relocation. According to the remote sensing image, high resolution can be found, man-made coastal reclamation also extends into the sea. It also exacerbated the regional topography changes.
According to the topographic effects on the "strange tide" occurred, establishes a model of trench velocity flux analysis by ArcGIS software. The terrain data by interpolation analysis, can be found in the northern Jiangsu shoal special topography into a plurality of sand ridges, to the coastal zone, the terrain will be shallow narrowing. According to the quantitative relationship between flow and velocity using MATLAB, EXCEL and other data programming software, obtained the quantitative velocity near the coast will increase the conclusion.
Using statistical methods to analyze marine element data, it is mainly divided into three models: Multivariable autoregressive model, artificial neural network model and nonlinear chaotic dynamic model.
Multivariate autoregressive model (AR model), is the ocean according to various elements of time scale arrangement, in accordance with the method of time series, to find the most suitable model order. It also obtains the effect of numerical ocean elements current time scale. According to the optimal order of the model, calculate the autocorrelation coefficient the model using the maximum likelihood method to establish AR model. Correlation coefficient matrix model, fully reflects the linear quantitative relationship between each other in different time scales of various marine elements. To find the reason velocity mutation through statistical methods. During the study period in this paper, the wind speed will have a certain effect of tide, tidal level will cause the increase or decrease, while the tidal changes will affect ocean velocity.
The artificial neural network model is designed, through the establishment of time delay network, find out the nonlinear relationship between the various marine elements in different time dimensions between each other. The characteristics of neural network is calculated in every neurons are very simple, but every single neuron layer after layer connected form the neural network is large and complex, like the brain. In the process of data fitting, the input vector variables using neural network fitting, the effect is not good, as the input vector of single variable fitting effect is good. The neural network model to predict the effect is not ideal, this is mainly because of too the weight number in the network, so the prediction of the results will have a great deviation.
Nonlinear chaotic dynamics model, and find out the nonlinear relationship between various marine elements. Another important function is to calculate the analysis data of the Lyapunov exponent and Kolmogrov entropy. According to the trusted time of these two indicators can be obtained to predict the scale. In this study, directions are discussed respectively. The wind speed East-West the tide and velocity, fitting and prediction of the effect is good, and credible time scale prediction is larger; and to the north and south wind, tidal flow and credible time scale is relatively short, chaotic phenomenon occurs more easily.

【学位授予单位】：上海交通大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：U698.6

【相似文献】