深圳大鹏湾波浪数值预报研究

发布时间：2018-03-13 19:22

本文选题：非线性自回归神经网络　切入点：WRF大气模式　出处：《清华大学》2015年硕士论文　论文类型：学位论文

【摘要】：深圳大鹏湾位于南海北部,西邻珠江口,东接大鹏半岛,沿岸建有港口、海水泳场、浅海养殖区、以及深水航道,人工作业繁忙。因此湾内波浪的准确预报对于大鹏湾内作业安全、以及海洋灾害预警具有十分重要的意义。同时深圳市为社会经济发展以及公众安全,在深圳周边海域建立了全面的海洋环境浮标监测网络。本文在该监测网络的基础上开展了波浪在线预报模型的研究,该模型包括两部分:基于数据驱动模式的浮标在线预报模型,以及基于海-气耦合数值模式的波浪场区域预报模型。首先基于大鹏湾波浪浮标在线监测数据,采用了数据驱动模式实现大鹏湾单点波浪的在线实时预报。采用两种非线性自回归人工神经网络NAR、NARX对大鹏湾内湾口、大梅沙与东涌浮标站分别进行了3h、6h以及12h的有效波高预报。结果显示大梅沙与湾口站点3h与6h预报的平均误差在0.1m以内;东涌3h、6h与12h预报误差均在0.1m以内,预报相关系数在0.904以上。基于大气中尺度模式WRF与浅水波浪数值模式SWAN的耦合模型WRF_SWAN建立了大鹏湾波浪场区域预报模式。并研究了WRF模式中不同参数如垂直层数、顶部气压、微物理过程对模拟结果的影响,以及采用地理信息管理GIS空间分析模型研究了土地覆被的变化对大鹏湾风场的影响;此外,研究了SWAN模式中白冠破碎速率对波浪场的影响。模型参数率定后大鹏湾内浮标站点的有效波高计算结果平均误差在0.29m以内。为验证WRF_SWAN耦合模式的适应性,将其应用于2014年第15号台风“海鸥”袭击大鹏湾期间的波浪场进行后报验证;验证结果表明在台风“海鸥”期间大鹏湾的有效波高后报平均误差在0.32m以内。最后采用NOAA的全球预报系统GFS提供的气象预报数据作为SWAN_WRF模式的预报初始与边界条件对大鹏湾波浪进行72h预报。湾口波浪预报绝对误差为0.23m,精度满足日常预报要求。计算结果说明本文所建的WRF_SWAN耦合模式适用于深圳大鹏湾的波浪预报。
[Abstract]:Dapeng Bay in Shenzhen is located in the northern part of the South China Sea, adjacent to the Pearl River Estuary to the west, and to the Dapeng Peninsula to the east. There are ports along the coast, sea water swimming fields, shallow water culture zones, and deep water channels. Therefore, accurate prediction of waves in the Bay is of great significance to operational safety in Mirs Bay and early warning of marine disasters. At the same time, Shenzhen City is responsible for social and economic development and public safety. A comprehensive marine environment buoy monitoring network has been established in the surrounding sea area of Shenzhen. Based on the monitoring network, the on-line wave prediction model has been studied in this paper. The model consists of two parts: the on-line prediction model of buoy based on data-driven model, and the prediction model of wave field based on coupled ocean-atmosphere numerical model. Firstly, the on-line monitoring data of wave buoy in Dapeng Bay are used. In this paper, a data-driven model is used to realize on-line real-time prediction of single point waves in Dapeng Bay. Two kinds of nonlinear autoregressive artificial neural networks (NARN NARX) are used to predict the mouth of Dapeng Bay. The effective wave height was predicted for 3 h and 12 h at Dameisha and Tung Chung buoy stations, respectively. The results show that the average error of 3 h and 6 h prediction for Dameisha and Bay mouth stations is less than 0.1 m, and the prediction errors for 3 h and 12 h in Tung Chung are within 0.1 m, respectively. The prediction correlation coefficient is more than 0.904. Based on the coupling model WRF_SWAN of mesoscale model WRF and shallow water wave numerical model SWAN, the regional prediction model of Mirs Bay wave field is established. The different parameters of WRF model, such as vertical layer number, top pressure, are studied. The effects of microphysical processes on simulation results and the effects of land cover changes on wind fields in Dapeng Bay are studied by using the GIS spatial analysis model of geographic information management. The effect of the breakup rate of the white crown on the wave field in the SWAN model is studied. The average error of the effective wave height calculated at the buoy station in Dapeng Bay is less than 0.29 m after the model parameter rate is fixed. In order to verify the adaptability of the WRF_SWAN coupling model, It was applied to the wave field during typhoon no. 15, 2014, when seagull attacked Mirs Bay. The results show that the average error of effective wave height in Dapeng Bay during typhoon "seagull" is less than 0.32 m. Finally, the meteorological forecast data provided by the global forecasting system (GFS) of NOAA are used as the initial prediction and boundary strip of SWAN_WRF model. The absolute error of wave prediction at the mouth of Dapeng Bay is 0.23m, which satisfies the requirement of daily forecast. The results show that the coupled WRF_SWAN model is suitable for the wave prediction of Dapeng Bay in Shenzhen.
【学位授予单位】：清华大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：P731.22

【相似文献】