PREDICTION OF TYPHOON SWELLS USING NEURAL ETWORKS 蕭松山 1 丁肇隆 2 林銘崇 2 蘇昭安 3  國立台灣海洋大學河海工程學系副教授  國立台灣大學工程科學及海洋工程學系教授  國立台灣大學工程科學及海洋工程學系碩士 指導老師 : 陳榮昌 老師.

Slides:



Advertisements
Similar presentations
控制原理與設計期中報告 指導教授:曾慶耀 學 號: 學 生:楊長諺.  Introduction  System Modeling of the PMAC Motor  Neural - Network - Based Self - Tuning PI Control System.
Advertisements

Wireless Power Transmission and Near Field Communication
-Artificial Neural Network- Chapter 2 Basic Model
Intelligent Database Systems Lab 國立雲林科技大學 National Yunlin University of Science and Technology 1 A 24-h forecast of solar irradiance using artificial neural.
Hazırlayan NEURAL NETWORKS Least Squares Estimation PROF. DR. YUSUF OYSAL.
Web-based Spoken English Training System with American Accent 線上美式英文口音訓練系統 指導教授:陳恆佑老師 學生:王舜霈 Date: June 25th, 2008 國立暨南國際大學 資訊工程學系碩士班畢業成果展 1.
國立雲林科技大學 National Yunlin University of Science and Technology Predicting adequacy of vancomycin regimens: A learning-based classification approach to improving.
-Artificial Neural Network- Counter Propagation Network
1 The numerical methods for Helmholtz equation 報告人:陳義麟 國立高雄海洋技術學院造船系副教授 於海洋大學河海工程系 基隆 2003/10/23.
12.4 Notes Weather Analysis
Derivation of stiffness and flexibility for rods and beams by using dual integral equations 海洋大學河海工程學系 報 告 者:謝正昌 指導教授:陳正宗 特聘教授 日期: 2006/04/01 中工論文競賽 (
楊竹星 國立成功大學電機工程系 98學年第一學期
國立台灣海洋大學 Partial differential equation 作業 4 教授:陳正宗終生特聘教授 學生:徐胤承 學號: M
劉金源、李澤民 國立中山大學海洋科學學院 海下科技暨應用海洋物理研究所教授、海洋生物研究所教授 海洋科學 Marine Sciences.
Spurious Eigensolutions and Fictitious Frequencies for Acoustic Problems with the Mixed-type Boundary Conditions by using BEM 邊界元素法於混合型邊界條件問題之 假根及虛擬頻率探討.
Implementing a reliable neuro-classifier
A new predictive search area approach for fast block motion estimation Kuo-Liang Chung ( 鍾國亮 ) Lung-Chun Chang ( 張隆君 ) 國立台灣科技大學資訊工程系暨研究所 IEEE TRANSACTIONS.
User Centered & Universal Design 1-7 國際文化交流課程發展計畫 針對研一及碩專班一年級學生 一學分 /18 小時.
Introduction to Environmental Engineering- Introduction 朱信 國立成功大學環境工程學系 環境工程概論 - 導論.
-Artificial Neural Network- Chapter 5 Back Propagation Network
A Simulation Study of the PWM Strategy for Inverters
國立陽明大學生資學程 陳虹瑋. Genetic Algorithm Background Fitness function ……. population selection Cross over mutation Fitness values Random cross over.
國立台灣海洋大學河海工程 研究所 BEM2004 第 8 次作業 博三 錢榮芳 D 博一 周家慶 D 碩一 吳安傑 M 碩一 李文愷 M Filename: BEM ppt by A. C. Wu 單雙層解法.
-Artificial Neural Network- Chapter 3 Perceptron 朝陽科技大學 資訊管理系 李麗華教授.
直銷通路管理報告 指導教授﹕陳得發教授 學生 : 王昭雄 學號﹕ 中華民國九十一年三月十三日 The Effect of Goal-Setting on the Performance of Independent Sales Agents in Direct Selling by Thomas.
南榮技術學院 101 學年度第二學期教師教學媒體觀摩及數位教材製作2015/9/11 工程科技研究所 副教授 蔡錦山 1 報告人:工程科技研究所 副教授 蔡錦山 A Study on the Linkage Mechanisms.
Design of double- and triple-sampling X-bar control charts using genetic algorithms 指導教授: 童超塵 作者: D. HE, A. GRIGORYAN and M. SIGH 主講人:張怡笳.
Great New England Hurricane Nicole Hartford.  The hurricane formed south of the Cape Verde Islands on September 9,  At this time, New England.
國立屏東商業技術學院 資訊工程系 ( 所 ) 多媒體技術發展實驗室 Laboratory of Multimedia Technology Development Department of Computer Science and Information Engineering Nation Pingtung.
Copyright © 2012, A Minimum Cost Resource Allocation Approach for Cloud Data Centers 指導教授:王國禎 學生:連懷恩 國立交通大學資訊工程系 行動計算與寬頻網路實驗室 1.
-Artificial Neural Network- Chapter 9 Self Organization Map(SOM) 朝陽科技大學 資訊管理系 李麗華 教授.
Intelligent Database Systems Lab N.Y.U.S.T. I. M. Comparison of neural network models with ARIMA and regression models for prediction of Houston's daily.
Department of Electrical Engineering Southern Taiwan University of Science and Technology Robot and Servo Drive Lab. 學生 : 蔡景棠 指導教授 : 王明賢 2015/10/13 A Driver.
98/02 國立台北科技大學能源與冷凍空調 工程研究所 施陽正 老師 1 高等熱傳學 (Advanced Heat Transfer) 能源與冷凍空調工程研究所 九十八年二月.
自我介紹 台中人 國立大里高中 交大運管系 黃明居老師 lab 2012/1/5 運輸組織與管理 1.
1 1 Institute of Industrial Engineering & Management National Yunlin University of Science & Technology Jen-Hao Lo Advisor: Chwen-Tzeng Su, Ph.D. Expansion.
Grid Programming on Taiwan Unigrid Platform. Outline Introduction to Taiwan Unigrid How to use Taiwan Unigrid.
Caribbean Disaster Mitigation Project Caribbean Institute for Meteorology and Hydrology Tropical Cyclones Characteristics and Forecasting Horace H. P.
An Assessment of The Effect of Contraflow Bus Lane on Rider Ship 指導教授:任維廉 葉家銘 Luke Yeh 1.
數量方法 課程名稱數量方法 課程編碼 60M01701 系所代碼 / 名稱 06 / 國企系 開課班級碩研國企一甲 開課教師林士琪 學分 3.0 時數 3 必選修系定選修 南台科技大學 課程資訊.
台灣師範大學機電科技學系 C. R. Yang, NTNU MT -1- Chapter 11 Numerical Integration Methods in Vibration Analysis 11.
科技素養教育 Technology Literacy 方榮爵 教授 南台科技大學 講座教授. What is Technological Literacy? Technological literacy is the ability to use, manage, assess, and understand.
Intelligent Database Systems Lab 國立雲林科技大學 National Yunlin University of Science and Technology Efficient Optimal Linear Boosting of a Pair of Classifiers.
Intelligent Database Systems Lab N.Y.U.S.T. I. M. The application of SOM as a decision support tool to identify AACSB peer schools Presenter : Chun-Ping.
Maximization of System Lifetime for Data-Centric Wireless Sensor Networks 指導教授:林永松 博士 具資料集縮能力無線感測網路 系統生命週期之最大化 研究生:郭文政 國立臺灣大學資訊管理學研究所碩士論文審查 民國 95 年 7 月.
All-to-all broadcast problems on Cartesian product graphs Jen-Chun Lin 林仁俊 指導教授:郭大衛教授 國立東華大學 應用數學系碩士班.
MethodFichera’s method The boundary flux equilibrium The CHEEF method The hypersingular formulation The method of adding a rigid body mode Formulation.
1 Vision for Marine and Coastal Services Digital Products Jamie Vavra Marine and Coastal Weather Services Branch Office of Climate, Water and Weather Services.
Research Writing Proposal The China Airline’s Strategy For Cross-strait flight policy By Jin Yang.
Copyright © 2011, A Resource Allocation Mechanism of Data Center for Public Cloud Service 指導教授:王國禎 學生:連懷恩 國立交通大學網路工程研究所 行動計算與寬頻網路實驗室.
Weather effects on the returns and volatility of the Shanghai stock market 國際財管 指導老師 : 何啟銘 學生 : 林武義 學號 :ma
A Case for Analytical Customer Management Group 林依祈 洪瑗伶 王裕濱 CRM paper reading presentation.
Stream Depletion Rate with Horizontal or Slanted wells in Confined Aquifers near a Stream 授課老師 : 陳文福 指導老師 : 馮正一 報告人 : 鄒佩蓉 1 探討水平井或傾斜井於河邊受壓含水層抽水之河川損耗率 (SDR)
南台科技大學 資訊工程系 An effective solution for trademark image retrieval by combining shape description and feature matching 指導教授:李育強 報告者 :楊智雁 日期 : 2010/08/27.
How are hurricanes formed ? A hurricane is a system of low pressure that originates over a tropical area. There are specific conditions that must be present.
國立交通大學生物科技學院 助理教授 林勇欣 一、參選類別(請勾選): ■「學士班必修課程與全校性共同課程」候選人 □「其他 ( 學士班選修及研究所 ) 課程」候選人 四、近二年任教之學士班必修課程與全校性共同課程.
輔仁大學 電機工程學系 大學部專題生 Investigate Partial CRC-32 Characteristic and Performance for Real-time Multimedia Streaming in Wireless Mesh Networks 指導教授:莊岳儒.
Ch.10 Conflict, Power, and Politics 報告學生:蘇世名 指導老師:任維廉 報告日期: 2009/12/ /3/171 國立交通大學 蘇世名.
微機電伺服控制實驗室 MEMS Servo Control System Lab. 國立交通大學機械工程學系工程五館 249 室 分機: 國立交通大學 機械工程學系 指導教授:陳宗麟 老師辦公位置:工程五館 469 室 (EE469) 電子信箱:
Joseph Fitzwater, Senior Analyzing Hurricane Intensity with a New Classification for the 21 st Century.
Accurate Robot Positioning using Corrective Learning
Wave height prediction in the Apostle Islands
Luís Filipe Martinsª, Fernando Netoª,b. 
Motivation Computers are good at some things… Calculating 
of the Artificial Neural Networks.
-Artificial Neural Network- Perceptron (感知器)
Yunxia Zheng, Yongping Li and Runling Yu OCT,2009 in Halifax
Error Minimization of Diffusion Operator
國立臺灣海洋大學河海工程研究所結構組 碩士班畢業論文口試
(海報論文:一般組或博士生組或碩士生組或大學生組)
Presentation transcript:

PREDICTION OF TYPHOON SWELLS USING NEURAL ETWORKS 蕭松山 1 丁肇隆 2 林銘崇 2 蘇昭安 3  國立台灣海洋大學河海工程學系副教授  國立台灣大學工程科學及海洋工程學系教授  國立台灣大學工程科學及海洋工程學系碩士 指導老師 : 陳榮昌 老師 工業工程與管理系 碩一 A 學生 : 林巧玲 、 白彩綾

Outline 1.Introduction 2.Back-Propagation Neural network (BPN) 3.Forming the BPN framework 4.Results and discussion 5.Conclusions 2

1.Introduction 3

Motivation The track of Typhoon Fred in 1994 did not hit Taiwan, however, the maximum significant wave height reached 7.59 m. The potential for such large waves cannot be ignored. The establishment of a swell warning system is now urgent. A precise warning system must yield good wave-height predictions. Accurately predicting swell motion is therefore extremely important. 4

Reference 5 World War Ⅱ Predictive models of wind waves have developed since World War II. These models predict the characteristics of waves from wind data. Sverdrup and Munk (1947) developed a model to predict significant wave heights generated by winds. Bretschneider (1952) used empirical data to present the well-known SMB method, which forecasts significant wind waves under steady-fetch conditions.

Reference 6 World War Ⅱ Wilson (1955) developed a graphical SMB method to overcome this weakness. Ijima et al. (1968) extended Wilson’s method to predict the heights of wind waves in hallow water. Based on Bretschneider’s method, Breschneider and Tamaye (1976) developed a method to forecast the heights of waves generated by hurricanes.

Objective Their numerical results did not agree closely with field measurements. The mechanism of swell generation by typhoons is extremely complicated, so no physical model is currently available to describe swell motion. A neural network method is used herein to improve predictions of typhoon swell. 7

Objective Neural networks are data-oriented, meaning that the relationship between the input and the output parameters is not necessarily assumed to be fixed. An ANN can tolerate errors in input data and still yield solutions. They can also adjust themselves effectively to new input data. 8

2.Back-propagation neural network BPN is a supervised learning model that mimics a human being’s nerve system. Learning from past records (experiences), BPN can optimally predict a target’s future behaviors. 9

2.Back-propagation neural network 10

11

12

3.Forming the BPN framework Fifteen sets of typhoon-related meteorological data and the corresponding time series of significant wave height data near Hualien and Suau Harbors. It provided by Central Weather Bureau in Taiwan, were used to establish the BPN framework and to confirm the ability of the network to forecast accurately. 13

3.Forming the BPN framework Define six parameters W max : the maximum speed of the wind storms. V : the speed of the typhoon. d : the distance from the center of typhoon to the observing station. R 7 : the radius of force 7 wind. θ 1 : the azimuthal angle. θ 2 : the angle between the direction of motion of the typhoon and the line that connects the center of the typhoon to the observing station. 14

3.Forming the BPN framework 15 root-mean-squared values (rms) of E.

3.Forming the BPN framework 16

3.Forming the BPN framework H p(t+1) is significant wave height predicted for the next hour at the observation station. 17

3.Forming the BPN framework 18 when η = 5 → 0.01 and α = 0.2

3.Forming the BPN framework 19

The mathematical expression of the BPN is 20 3.Forming the BPN framework

4. Results and discussion 21

22

23

24

25

26

27

28

5. Conclusions The back-propagation neural network (BPN), an artificial neural network, mimics the nerve system of a human being. Learning from the past records (history), BPN can predict the future behavior of an object. Several data (12 sets) concerning typhoons and the associated measured wave heights at Hualien were fed into BPN to determine the optimal relationship among elements in the BPN. An optimal framework of BPN was thus developed by trial and error. 29

5. Conclusions The developed BPN was tested by predicting several examples. The BPN was found to predict the measured data very accurately. However, as the predictions are made further into the future (such as over two hours or three hours), the forecast of the BPN becomes worse perhaps because the unexpected path of the typhoons cause inaccuracies. This problem can be solved by improving the forecasts of the characteristics of typhoons and training the BPN often with new typhoon swell data. The BPN method is generally suitable for forecasting swells. 30