摘  要
气象学上指海岸地区的一种风，在白天从海洋中吹到陆地，与陆上的风相反。陆风和海风属于温和的类型，所以又被称作海陆之风。海陆性质的差异，也是海陆清风的成因。在白天，由于阳光的照射，地面温度迅速上升，地面温度高于海洋，大气因热量的增加而膨胀，从而导致低气压下降，高气压增加；而在海上则是低气压上升，高气压下降。因此，在海陆交界处，大气的底层表面气压比陆地要高，如果不考虑其他因素，那么气流就会从高气压向低气压方向流动，也就是说，海风会从海洋表面吹到地面。通常情况下，海上的风只有4-5级。海风的范围并不大，水平方向最多也就是几十公里，高度也只有几十米，就算再大的海风，深入内陆不过50～60公里。海陆风虽然出现在海滨地区，但也不是所有海滨地区都有海陆风现象。在地区上，中、低纬度出现的机会多；从季节上看，夏天要比其它季节更突出。海洋和陆地的周期性的风，在一定程度上影响着沿海的气候。由于海风的吹拂，大量的海洋潮汐进入了陆地，在陆地上形成了一层薄雾，直至有了降雨，海岸的温度就下降了，夏天也就不那么炎热了，但是关于海风的分类和预测方面的论文很少，目前来说正处于探索和研发阶段。
本文参考众多国内外相关研究后，分析对比了卷积神经网络、长短期记忆(LONG SHORT-TERM MEMEORY,LSTM)和极限梯度提升（eXtreme GradientBoosting，XGBoost）模型在台风强度预测中的应用模型特点，并分析它们在台风强度预测应用中的利弊，提出了基于Django进行海风分类及预测的优化方法，使用卫星传输的气象数据进行实验，确定了优化模型的相关参数。为验证优化模型的可行性与优越性，设计了优化模型与原模型的多组对比分析实验，并选取训练数据集以外的代表性海风进行了预测实验。实验结果表明，优化模型应用于海风强度的预测是可行的，且在短时临近海风分类预报中，准确率更高、 稳定性更好。本文提出的优化模型在海风强度分类预测领域的应用研究，拓展了海风分类和预测方式，为深度学习的相关方法在气象领域的应用和发展提供了新思路。
关键词：海风；分类和预测；Django
ABSTRACT
In meteorology, a wind in coastal areas blowing from the ocean to land during the day as opposed to the wind on land. Land and sea breezes are mild, so they are called land and sea breezes. The difference of the nature of land and sea is also the cause of the sea breeze. During the day, as the sun shines, the temperature of the ground rises rapidly, the temperature of the ground is higher than that of the ocean, and the atmosphere expands as a result of the increase in heat, which causes the low pressure to fall and the high pressure to increase; At sea, low pressure rises and high pressure falls. Thus, at the boundary between land and sea, the lower surface pressure of the atmosphere is higher than that of the land, and if nothing else, air flows from high pressure to low pressure, that is, sea winds blow from the surface of the ocean to the ground. Normally, winds out at sea are only force 4-5. The range of sea breeze is not large, the horizontal direction is at most dozens of kilometers, the height is only dozens of meters, even if the biggest sea breeze, only 50 to 60 kilometers inland. Although land and sea wind appears in the coastal area, but not all coastal areas have the phenomenon of land and sea wind. Regionally, the middle and low latitudes have more opportunities; In terms of season, summer is more prominent than other seasons. The coastal climate is affected to some extent by periodic winds over the sea and land. Due to the wind blowing, a large amount of ocean tide enters the land, forming a haze on the land, until the rainfall, the temperature of the coast is lowered, and the summer is not so hot, but there are few papers on the classification and prediction of sea breeze, currently in the stage of exploration and development.
In this paper, convolutional neural network, LONG short-term MEMEORY (LSTM) and eXtreme GradientBoosting (LSTM) are analyzed and compared after referring to many relevant researches at home and abroad. The advantages and disadvantages of XGBoost (XGBoost) model in typhoon intensity prediction are analyzed, and an optimization method of sea breeze classification and prediction based on Django is proposed. Experiments are carried out using meteorological data transmitted by satellite, and the parameters of the optimization model are determined. In order to verify the feasibility and superiority of the optimized model, multiple comparative analysis experiments between the optimized model and the original model were designed, and representative sea breeze prediction experiments were conducted outside the training data set. The experimental results show that the optimized model is feasible to predict the sea breeze intensity, and it has higher accuracy and better stability in the short-time classification forecast of nearby sea breeze. The application research of the optimization model proposed in this paper in the classification and prediction of sea breeze intensity extends the classification and prediction methods of sea breeze, and provides a new idea for the application and development of deep learning related methods in the field of meteorology.
Keywords：Sea breeze; Classification and prediction; Django
目 录
摘  要 I
ABSTRACT IV
第一章 概述 1
1.1系统背景 1
1.2系统意义 1
1.3 国内外研究现状 2
1.4 系统所需的软件资源 2
1.4.1 系统开发环境 2
1.4.2  关于python介绍 2
1.4.3 关于djando介绍 2
1.4.4关于数据库技术介绍 2
第二章 系统的总体设计 4
2.1 目标设计 4
2.2 系统的可行性研究分析 4
2.3 系统的数据分析 4
2.4 系统的功能模块设计 4
第三章 系统分析 6
3.1 系统流程分析 6
3.1.1 系统开发流程 6
3.1.2 用户登录流程 6
3.1.3 系统操作流程 7
3.1.4 添加信息流程 8
3.1.5 修改信息流程 8
3.1.6 删除信息流程 9
3.2 系统分析 9
3.2.1 系统的工作原理 9
3.2.2 系统结构设计 10
3.2.2 系统设计原则 11
3.3  系统实施分析 11
3.4  系统维护分析 12
数据库表结构 13
第四章 系统实现 14
4.1登录 14
4.2首页 14
第五章 系统的测试运行 15
5.1测试环境 15
5.2 测试目的 15
5.3 测试计划 16
5.4 测试用例 16
5.5 压力测试 18
5.6 测试小结 19
总 结 20
参考文献 21
致谢 22