基於改進的趨勢分類模型運用於天氣預測｜國立臺灣科技大學博碩士論文系統

簡易檢索 / 詳目顯示

回結果列表

研究生：	章陽中 Yang-Chung - Chang
論文名稱：	基於改進的趨勢分類模型運用於天氣預測 Trend Classification Based Weather Forecast
指導教授：	蘇順豐 Shun-Feng Su
口試委員:	李祖添 Tsu-Tian Lee 莊鎮嘉 Chen-Chia Chuang 王乃堅 Nai-Jian Wang
學位類別：	碩士 Master
系所名稱：	電資學院 - 電機工程系 Department of Electrical Engineering
論文出版年：	2017
畢業學年度：	105
語文別：	英文
論文頁數：	126
中文關鍵詞：	趨勢、分類模式、大數據、氣象局、天氣預測、卡爾曼濾波器
外文關鍵詞：	Trend, Classification mode, Big data, Bureau of Meteorology, Weather forecast, Kalman Filter
相關次數：	點閱：327 下載：13
分享至:	分享至facebook 分享至twitter

查詢本校圖書館目錄查詢臺灣博碩士論文知識加值系統勘誤回報

此研究貢獻在於提出一個有別於傳統技術的趨勢分類模型的新型天氣預報：
運用溫度歷史資料於未來溫度預測，以台北為例。由於考慮的數據量是很巨大的，
所提出的方法也可以被視為大數據資料分析，即為：使用數據分析解決大量數據。
方法是運用溫度的歷史資料作為資料庫並透過資料分析、建模來預測氣溫。通常，
每日溫度以區間間隔於建模，因此，在此研究的方法中，上限和下限（通常表示
當日最高溫、最低溫）是被考慮用於匹配／預測，而趨勢分類是提供一個層次分
類結構的模型，先使用溫差趨勢概念再整合傳統分類方法，使此模型能將匹配點
更具意義，而基於這模式，將可使溫度預測模型成功被使用。而在趨勢分類模型
中，更考量重疊的概念以及數據相關的對應關係，使模式匹配更有意義。並在趨
勢分類後，使用聚類方法來定義所匹配的模式。考慮重疊的想法與簡潔，本研究
使用上限與下限做為資料的區隔，更以上限及下限表示一天的溫度型態。本研究
的方法是在進行模糊分類之前先做趨勢分析，而溫度的劇烈變化也被考慮並且在
模式定義過程中被視為異常值以避免模式誤差過大。在處理數據時，透過不同模
式將資料型態做分別不同的處理方式。從實驗結果得知，透過此研究的方法在預
測一天中的最高溫度之平均誤差與新聞預測(中央氣象局預測)相比，平均誤差至
少低0.5℃；最低溫度之平均誤差與新聞預測(中央氣象局預測)相比，平均誤差
至少低0.8℃。換句話說，此研究之預測結果與氣象局做出的傳統預測是一致的，
甚至更好。此外，透過數據多方考量的技術概念，本項研究使用靜態卡爾曼濾波
器，將我們方法的預測結果和來自氣象局預測的結果組合，從實驗結果可以明顯
看出，這種結合性的方法，實際上比單一個方法做出的預測結果會有更好現象。

This thesis reports our study on a novel weather forecast approach based on
trend classification. This approach is to predict future temperature based on past
temperature histories. Usually the daily temperature is modeled as an interval and
hence, in our approach, both the upper bound and lower bound are considered for
match/prediction. Since the amount of data considered is huge, the proposed approach
can also be viewed as big data analysis. In our approach, trend classification is to
provide a hierarchical clustering structure to make pattern match more meaningful so
that temperature prediction based on those patterns can work. In the proposed trend
classification, an overlapping idea is considered to cope with possible variances of
temperatures in trend classification. After trend classification, a clustering approach is
employed to define patterns for match. In this study, dramatic changes in temperatures
are also considered and are treated as outliers in the pattern definition process to avoid
pattern pollution. After the processing of data, the patterns are used to predict the
temperatures in the future. From the experimental results, the average error for the
upper bound temperature predicted for the following four days by the proposed
approach is at least 0.5C smaller than that reported the news, and the average error of
the lower bound temperature is at least 0.8C smaller than that in the news. In other
words, the prediction only using past data without any meteorological information is
compatible with and even better than the traditional prediction using meteorological
information. Moreover, a simple data fusion technique by using Kalman filter is also
considered to combine the predictions from our approach and from meteorological
information. From the experiments, it can be evident that such an approach can indeed
have much better prediction than those made by individual approaches.

中文摘要........................................................................................................................ I
Abstract ......................................................................................................................... II
Figure list ...................................................................................................................... V
Table list.................................................................................................................... VIII
Acknowledgements......................................................................................................IX
Chapter 1 Overview .......................................................................................................1
1 Motivation........................................................................................................1
2 Related work ....................................................................................................2
3 Research Objective ..........................................................................................6
4 Organization.....................................................................................................7
Chapter 2 Background Knowledge ................................................................................8
1 Neural network.................................................................................................8
1.1 Back-Propagation Neural Network...............................................................9
2 Correlation in cities..........................................................................................9
2.1 Linear regression.........................................................................................10
3 Delay Time.....................................................................................................12
3.1 Correlation coefficient ................................................................................12
3.2 Linear Interpolation Method .......................................................................13
3.3 Pearson correlation coefficient ...................................................................14
3.4 Discussions .................................................................................................15
4 Data pre-processing .......................................................................................16
4.1 Conception ..................................................................................................16
4.2 Data form and resolutions ...........................................................................17
4.3 Classification mode.....................................................................................17
4.3.1 Euclidean distance ...................................................................................18
4.3.2 Fuzzy c-means cluster..............................................................................19
4.4 Discussions .................................................................................................19
5 Remarks .........................................................................................................20
Chapter 3 Trend Classification.....................................................................................52
1 Concept ..........................................................................................................52
2 Methodology ..................................................................................................53
3 Data ................................................................................................................54
3.1 Interval data ................................................................................................55
3.2 Data form ....................................................................................................55
3.3 Input data ....................................................................................................56
3.4 Target Data ..................................................................................................56
IV
3.5 Resolution ...................................................................................................56
3.6 Severe situation ...........................................................................................57
3.7 Data shifts ...................................................................................................57
4 Trend ..............................................................................................................58
4.1 Three types of trend ....................................................................................58
4.2 Improved three types of trend .....................................................................58
4.3 Seven types of trend....................................................................................59
5 Classification..................................................................................................60
5.1 Trend Classification mode ..........................................................................60
6 Improved Trend Classification mode.............................................................61
7 Fuzzy c-mean .................................................................................................63
8 Euclidean distance .........................................................................................64
9 Remarks .........................................................................................................65
Chapter 4 Experimental results....................................................................................74
1 Forecast in each month ..................................................................................75
2 General case and ordinary case......................................................................79
3 Different resolutions ......................................................................................80
4 Remarks .........................................................................................................82
Chapter 5 Static state Kalman Filter ............................................................................91
1 General case ...................................................................................................91
2 Ordinary case .................................................................................................93
3 Combine ordinary and dramatic change case ................................................95
Chapter 6 Conclusions and Future Work ...................................................................106
1 Conclusions............................................................................................106
2 Future Work ...........................................................................................108
References.................................................................................................................. 110
                                

[1] H. Ong, C.-M. Wu, and H.-C. Kuo, “Hybrid Mass Flux Cumulus Schemes
(HYMACSs) as a step to unified cumulus parameterization and its application
to tropical cyclone intensity prediction,” in Proc. Weather AnalysisForecasting
Conference, vol. A2, pp. 2-13, 2015
[2] G.-D. Hwang, and J.-P. Chen, “A Two-Moment Cloud Microphysics
Parameterization in the Global Model,” in Proc. of Weather
AnalysisForecasting Conference, vol. P2, pp. 2-30, 2015.
[3] Charney, J. G., and A. Eliassen, “On the growth of the hurricane depression,” in
Proc. J. Atmos. Sci., vol. 21, pp. 68–75, Jan. 1964.
[4] Arakawa, A., and C.-M. Wu, “A unified representation of deep moist
convection in numerical modeling of the atmosphere,” in Proc. J.Atmos. Sci.,
vol. 70, pp. 1977-1992, Jul. 2013.
[5] A. Klein, R. S. Lee, and M. Robinson, “A refined model for identifying
unavoidable portion of the impact of weather and other factors on airport
operation,” in Proc. IEEE Digital Avionics Systems Conference, vol. 4B, pp. 1 -
13, Oct. 2011.
[6] Y. Wang, “Prediction of weather impacted airport capacity using RUC-2
forecast,” in Proc. IEEE Digital Avionics Systems Conference, vol. 3C, pp. 1 -
12, Oct. 2012.
[7] J. E. Young, A. Crowell, and A. Fabian, “Modeling weather in simulation and
analysis,” in Proc. Integrated Communications, Navigation and Surveillance
Conference (ICNS), pp. 1 - 11, Apr. 2013.
111
[8] P. Louka, G. Galanis, N. Siebert, G. Kariniotakis, P. Katsafados, I. Pytharoulis,
and K. G., “Improvements in wind speed forecasts for wind power prediction
purposes using Kalman filtering,” in. Proc. Journal of Wind Engineering and
Industrial Aerodynamics, vol. 96, no. 12, pp. 2348–2362, 2008.
[9] Wang, Y., and Kulkarni, D., “Modeling Weather Impact on Ground Delay
Programs,” in Proc. SAE Int. J. Aerosp., vol. 4, no. 2, p. 1207 - 1215, 2011.
[10] Z. Huang, and Y. Liu, “Combined use of double weather radar in weather
modification,” in Proc. International Conference on Information Science and
Engineering (ISISE), pp. 326 – 330, Dec. 2010.
[11] M. Zhua, Y. Pana, Z. Huangb, and P. Xua, “An alternative method to predict
future weather data for building energy demand simulation under global climate
change,” Energy and Buildings, vol. 113, pp. 74-86, Feb. 2016.
[12] H. S. Badr, B. M. Elhadidi, and A. O. Sherif, “Evaluation of data assimilation
on numerical weather prediction for Egypt,” in Proc. IEEE International on
Geoscience and Remote Sensing Symposium (IGARSS), 2010, pp. 3526-3529.
[13] M. Christensona, H. Manza, and D. Gyalistrasb, “Climate warming impact on
degree-days and building energy demand in Switzerland,” Energy Conversion
and Management, vol. 47, pp. 671-686, Apr. 2006.
[14] Mark F. Jentsch, AbuBakr S. Bahaj, and Patrick A.B. James, “Climate change
future proofing of buildings—Generation and assessment of building simulation
weather files,” Energy and Buildings, vol. 40, pp. 2148-2168, Fun. 2008.
[15] A. Budiansyah, and V. Mutiawani , “Building weather season pattern from aceh
local news article (Text period analysis via significant term),” in Proc.
International Conference on Information Technology Systems and Innovation
(ICITSI), pp. 29-32, Nov. 2014
112
[16] L. D. Aponte-Bermúdez J. M. Morell, and E. Rodríguez, “CariCOOS:
Improving high-resolution numerical weather prediction for the northeast
caribbean region,” in Proc. IEEE Oceans, pp. 1-10, Oct. 2015.
[17] J. D. Wang, “An external memory approach to computing the maximal repeats
across classes of DNA sequences,” in Proc. Asian J. Health Inf. Sci., vol. 1, pp.
276–295, 2006.
[18] S. F. Su, C. C. Chuang, C. W. Tao, J. T. Jeng, and C. C. Hsiao, “Radial Basis
Function Networks With Linear Interval Regression Weights for Symbolic
Interval Data,” in Proc. IEEE Transactions on Systems, Man, and Cybernetics,
Part B, vol. 42, pp. 69-80, Feb. 2012
[19] C. C. Chuang, C. W. Li, C. C. Hsiao, S. F. Su, and Jin-Tsong Jeng, “Robust
interval support vector interval regression networks for interval-valued data
with outliers,” in Proc. 7th International Conference on Soft Computing and
Intelligent Systems (SCIS) and 15th International Symposium on Advanced
Intelligent Systems (ISIS), Dec. 2014.
[20] M. A. O. Domingues, R. M. C. R. de Souza, and F. J. A. Cysneiros, “Arobust
method for linear regression of symbolic interval data,” Pattern Recognit. Lett.,
vol. 31, no. 13, pp. 1991–1996, Oct. 2010.
[21] C. C. Chuang, “Extended support vector interval regression networks for
interval input-output data,” Inf. Sci., vol. 178, no. 3, pp. 871–891, Feb. 2008
[22] Francisco de A.T. de Carvalho, Renata M.C.R. de Souza, Marie Chavent and
Yves Lechevallier, “Adaptive Hausdorff distances and dynamic clustering of
symbolic interval data,” Pattern Recognition Letters, vol. 27, pp. 167-179,
2006.
113
[23] Gang Xiang, Scott A. Starks, Vladik Kreinovich, and Luc Longpre, “New
Algorithms for Statistical Analysis of Interval Data,” Proceedings of the
Workshop on State-of-the-Art in Scientific Computing PARA'04, Denmark,
2004.
[24] D. D. Nguyena, L. T. Ngoa, L. T. Phama, and W. Pedryczb, “Towards hybrid
clustering approach to data classification: Multiple kernels based
interval-valued Fuzzy C-Means algorithms,” Fuzzy Sets and Systems, vol. 279,
pp. 17-39, Nov. 2015.
[25] L. Zhanga, W. Lua, X. Liua, W. Pedryczc, and C. Zhonga, “Fuzzy C-Means
clustering of incomplete data based on probabilistic information granules of
missing values,” Knowledge-Based Systems, vol. 99, pp. 51-70, May. 2016.
[26] H. Izakiana, and W. Pedrycza, “Agreement-based fuzzy C-means for clustering
data with blocks of features,” Neurocomputing, vol. 127, pp. 266-280, Mar.
2014.
[27] H. Tanaka, S. Vejima, and K. Asai, “Linear regression analysis with fuzzy
model,” in Proc. IEEE Transactions on Systems, Man, and Cybernetics., vol. 12,
no. 6, pp. 903–907, Nov. 1982.
[28] B. B. Chaudhuri and A. Rosenfeld, ”A modified Hausdorff distance between
fuzzy sets,” Information Sciences, vol. 118, pp. 159-171, 1999.

簡易檢索 / 詳目顯示

相關論文