臺灣博碩士論文加值系統

每個地區的主要汙染物可能會因為其氣候條件、地理位置與工業發展的程度而有所差異，如何準確預測未來的空氣品質使人們可以提前採取自我保護的措施，是許多研究人員的目標。
本論文使用長短期記憶(Long Short-Term Memory, LSTM)網路以每小時為單位，對台灣歷年來最主要的三種汙染物：細懸浮微粒(PM2.5)、臭氧(O3)與懸浮微粒(PM10)，進行跨小時的直接預測(Direct Forecast, DF)與只能預測下一小時的更新預測(Update Forecast, UF)。不同於多數研究預測的目標為汙染物的濃度值，本研究預測的目標是大眾較為熟知的空氣品質指標值(Air Quality Index, AQI)，本研究將LSTM的預測值結合模糊演算法提出一種新的綜合評估指標(LSTM-Fuzzy)，讓民眾可以更方便的了解未來的空氣品質概況。
實驗結果得知，直接預測的準確度通常會隨著預測時間的增加而降低，然而更新預測的準確度則相對於直接預測平穩，以12小時的預測結果來說，在四種監測站與四種季節的各式情況下的均方根誤差(Root Mean Square Error, RMSE)或平均絕對誤差(Mean Absolute Error, MAE)可以有低於10、甚至低於5的能力，其中只有夏季-交通站的預測能力較為不理想，其RMSE為19.96、MAE為15.42。本論文所提出的LSTM-Fuzzy準確率，在四種季節與四種監測站情況下12小時以內的預測準確率皆為100%，而夏季、秋季、冬季則是16小時以內會有100%的準確率。

The main pollutants in each area may vary due to their climatic conditions, geographic location and the development of industry. How to accurately predict the future air quality so that people can take self-protection measures in advance is the goal of many researchers.
In this paper, the Long Short-Term Memory (LSTM) network is used to measure the three major pollutants in Taiwan over the years: PM2.5, O3 and PM10, perform the cross-hour Direct Forecast (DF) and Update Forecast (UF) that can only predict the next hour. Unlike the target predicted by most studies for the concentration of pollutants, this study focus on the Air Quality Index (AQI), which is well known to the public. A new comprehensive evaluation method is proposed by combining the predicted value of LSTM with the fuzzy algorithm (LSTM-Fuzzy), so that the public can understand the future air quality profile more easily.
The experimental results show that the accuracy of DF usually decreases with the increase of the prediction time, but the accuracy of UF is relatively stable compared to the DF. Root Mean Square Error (RMSE) or Mean Absolute Error (MAE) can be less than 10 or even less than 5 in all kinds of seasons, only summer-traffic air quality monitoring stations the prediction ability is relatively unsatisfactory, with an RMSE of 19.96 and a MAE of 15.42. The accuracy rate of the proposed LSTM-Fuzzy is 100% within 12 hours under four seasons and four monitoring stations, while 100% accuracy within 16 hours in summer, autumn and winter.

目 錄
論文審定書
摘 要
Abstract
誌 謝
目 錄
圖 目 錄
表 目 錄
第1章 緒論
1.1 研究背景
1.2 研究動機
1.3 研究目的
1.4 研究範圍與限制
1.5 論文架構
第2章 文獻探討
2.1 空氣汙染物對人體的危害
2.2 台灣空氣品質歷史概況
2.2.1 台灣AQI不良天數佔比
2.2.2 台灣空氣品質監測站種類
2.3 高雄市的地理位置與歷史背景
2.4 高雄市空氣品質概況
2.4.1 高雄市空氣汙染源
2.4.2 高雄市空氣品質監測站
2.5 相關預測模型
2.5.1 確定性方法
2.5.2 統計方法
2.5.3 機器學習方法
第3章 相關理論基礎
3.1 深度學習概述
3.1.1 激活函數
3.1.2 梯度最佳化相關方法
3.1.3 Adam梯度最佳化方法
3.2 循環神經網路
3.3 長短期記憶網路
3.4 模糊演算法概述
3.4.1 模糊化與歸屬函數
3.4.2 模糊規則庫與模糊運算
3.4.3 模糊推論與解模糊化
第4章 研究方法與系統設計
4.1 系統架構
4.2 系統軟體設計與構成
4.2.1 Matlab編譯軟體
4.2.2 手機APP開發軟體
4.3 研究流程
4.3.1 數據集
4.3.2 資料預處理
4.3.3 LSTM網路參數設置
4.3.4 模糊系統之設計
4.3.5 手機顯示結果
第5章 實驗結果與討論
5.1 實驗環境
5.2 實驗結果與分析
5.2.1 LSTM評量指標
5.2.2 一年數據集與三年數據集
5.2.3 春季
5.2.4 夏季
5.2.5 秋季
5.2.6 冬季
5.2.7 LSTM-Fuzzy與APP介面
5.2.8 LSTM-Fuzzy的準確率
5.3 討論
第6章 結論與未來展望
6.1 結論
6.2 未來展望
參考文獻

[1]World Health Organization. Available at:
https://www.who.int/news-room/feature-stories/ten-threats-to-global-health-in-2019. (Accessed on 30 June 2020)
[2]World Health Organization. Available at:
https://www.who.int/news-room/detail/02-05-2018-9-out-of-10-people-worldwide-breathe-polluted-air-but-more-countries-are-taking-action. (Accessed on 30 June 2020)
[3]World Health Organization. Available at:
https://www.who.int/news-room/detail/29-10-2018-more-than-90-of-the-world%E2%80%99s-children-breathe-toxic-air-every-day. (Accessed on 30 June 2020)
[4]Environmental Protection Administration Executive Yuan, R.O.C.
Available at:
https://airtw.epa.gov.tw/CHT/Information/Standard/AirQualityIndicator.aspx. (Accessed on 30 June 2020)
[5]B.J. Lee, B. Kin, K. Lee, 2014, “Air Pollution Exposure and Cardiovascular Disease”, Toxicological Research, Vol. 30, Issue 2, pp. 71-75.
[6]S. Feng, D. Gao, F. Liao, F. Zhou, X. Wang, 2016, “The health effects of ambient PM2.5 and potential mechanisms”, Ecotoxicology and Environmental Safety, vol. 128, pp. 67-74.
[7]O. Raaschou-Nielsen, R. Beelen, M. Wang, G. Hoek, Z.J. Andersen, B. Hoffmann, M. Stafoggia, E. Samoli, G. Weinmayr, K. Dimakopoulou, M. Nieuwenhuijsen, W.W. Xun, P. Fischer, K.T. Eriksen, M. Sørensen, A. Tjønneland, F. Ricceri m, K. de Hooghk, T. Key, M. Eeftens, P.H. Peeters,H.B. Bueno-de-Mesquita, K. Meliefste, B. Oftedal, P.E. Schwarze, P. Nafstad, C. Galassi,E. Migliore, A. Ranzi, G. Cesaroni, C. Badaloni, F. Forastiere, J. Penell, U. De Faire, M. Korek,N. Pedersen, C.-G. Östenson, G. Pershagen, L. Fratiglioni, H. Concin, G. Nagel, A. Jaensch,A. Ineichen, A. Naccarati, M. Katsoulis, A. Trichpoulou, M. Keuken, A. Jedynska, I.M. Kooter,J. Kukkonen, B. Brunekreef, R.S. Sokhi, K. Katsouyanni, P. Vineis, 2016, “Particulate matter air pollution components and risk for lung cancer”, Environment International, vol. 87, pp. 66-73.
[8]Environmental Protection Administration Executive Yuan, R.O.C.
Available at:
https://iaq.epa.gov.tw/indoorair/page/News_7_6.aspx.
(Accessed on 30 June 2020)
[9]M. Jerrett, R. T. Burnett, C. Arden Pope III, K. Ito, G. Thurston, D. Krewski, Y. Shi, E. Calle, M. Thun, 2009, “Long-term ozone exposure and mortality”, N Engl J Med, 360(2009), pp. 1085-1095.
[10]J. D. Stowella, Y. Kimb, Y. Gaoc, J. S. Fud, H. H. Change, Y. Liua, 2017, “The impact of climate change and emissions control on future ozone levels: Implications for human health”, Environment International, vol. 108, pp. 41-50.
[11]Y. O. Khaniabadi, S. M. Daryanoosh, P. K. Hopke, M. Ferrante, A. D. Marco, P. Sicard, G. O. Conti, G. Goudarzi, H. Basiri, M. J. Mohammadi, F. Keishams, 2017, “Acute myocardial infarction and COPD attributed to ambient SO2 in Iran”, Environmental Research, vol. 156, pp. 683-687.
[12]J. C. Uzoigwe, T. Prum, E. Bresnahan, M. Garelnabi, 2013, “The Emerging Role of Outdoor and Indoor Air Pollution in Cardiovascular Disease”, N Am J Med Sci, vol. 5(8), pp. 445-453.
[13]H. Brønnum-Hansen, A. M. Bender, Z. J. Andersen, J. Sørensen, J. H. Bønløkke, H. Boshuizen, T. Becker, F. Diderichsen, S. Loft, 2018, “Assessment of impact of traffic-related air pollution on morbidity and mortality in Copenhagen Municipality and the health gain of reduced exposure”, Environment International, vol. 121, pp. 973-980.
[14]Environmental Protection Administration Executive Yuan, R.O.C.
Available at:
https://taqm.epa.gov.tw/taqm/tw/AqiOverMonthlyReport.aspx. (Accessed on 25 February 2020)
[15]Environmental Protection Administration Executive Yuan, R.O.C.
Available at:
https://airtw.epa.gov.tw/CHT/EnvMonitoring/Central/Background_Intro.aspx. (Accessed on 30 June 2020)
[16]Environmental Protection Administration Executive Yuan, R.O.C.
Available at:
https://erdb.epa.gov.tw/Subjects/MetaSubject.aspx?topic1=%E5%A4%A7%E6%B0%A3&topic2=%E7%92%B0%E5%A2%83%E5%8F%8A%E7%94%9F%E6%85%8B%E7%9B%A3%E6%B8%AC&subject=%E7%A9%BA%E6%B0%A3%E5%93%81%E8%B3%AA. (Accessed on 30 June 2020)
[17]Environmental Protection Administration Executive Yuan, R.O.C.
Available at:
https://erdb.epa.gov.tw/Subjects/MetaSubject.aspx?topic1=%E5%85%B6%E4%BB%96&topic2=%E7%92%B0%E4%BF%9D%E7%B5%B1%E8%A8%88&subject=%E7%92%B0%E4%BF%9D%E7%A4%BE%E7%B6%93%E7%B5%B1%E8%A8%88.
(Accessed on 30 June 2020)
[18]Dept. of Household Registration. Available at:
https://www.ris.gov.tw/app/portal/346. (Accessed on 30 June 2020)
[19]Kaohsiung Motor Vehicles Office Directorate General of Highways. Available at:
https://thbdemo.thb.gov.tw/komv/modules/govinfo/govinfo_list?node=7578f6a6-f0d0-41ff-90cc-2934a1d91da2&c=6.
(Accessed on 30 June 2020)
[20]P. E. Saide, G. R. Carmichael, S. N. Spak, L. Gallardo, A. E. Osses, M. A. Mena-Carrasco, M. Pagowski, 2011, “Forecasting urban PM10 and PM2.5 pollution episodes in very stable nocturnal conditions and complex terrain using WRF–Chem CO tracer model”, Atmospheric Environment, vol. 45, pp. 2769-2780.
[21]R. Sterna, P. Builtjes, M. Schaap, R. Timmermans, R. Vautard, A. Hodzic, M. Memmesheimer, H. Feldmann, E. Renner, R. Wolke, A. Kerschbaumer, 2008, “A model inter-comparison study focussing on episodes with elevated PM10 concentrations”, Atmospheric Environment, vol. 42, pp. 4567-4588.
[22]J. Chen, J. Lu, J. C. Avise, J. A. DaMassa, M. J. Kleeman, A. P. Kaduwela, 2014, “Seasonal modeling of PM2.5 in California's San Joaquin Valley”, Atmospheric Environment, vol. 92, pp. 182-190.
[23]K. Vijayaraghavan, S. Cho, R. Morris, D. Spink, J. Jung, R. Pauls, K. Duffett, 2016, “Photochemical model evaluation of the ground-level ozone impacts on ambient air quality and vegetation health in the Alberta oil sands region: Using present and future emission scenarios”, Atmospheric Environment, vol. 141, pp. 209-218.
[24]L. Jian, Y. Zhao, Y. P. Zhu, M. B. Zhang, D. Bertolatti, 2012, “An application of ARIMA model to predict submicron particle concentrations from meteorological factors at a busy roadside in Hangzhou, China”, Science of The Total Environment, vol. 426, pp. 336-345.
[25]C. Li, N. C. Hsu, S.C. Tsay, 2011, “A study on the potential applications of satellite data in air quality monitoring and forecasting”, Atmospheric Environment, vol. 45, pp. 3663-3675.
[26]P.J. García Nieto, E.F. Combarro, J.J. del Coz Díaz, E. Montañés, 2013, “A SVM-based regression model to study the air quality at local scale in Oviedo urban area (Northern Spain): A case study”, Applied Mathematics and Computation, vol. 219, pp. 8923-8937.
[27]J. M. Davis, P. Speckman, 1999, “A model for predicting maximum and 8 h average ozone in Houston”, Atmospheric Environment, vol. 33, pp. 2487-2500.
[28]X. Hu, L. A. Waller, M. Z. Al-Hamdan, W. L. Crosson, M. G. Estes Jr, S. M. Estes, D. A. Quattrochi, J. A. Sarnat, Y. Liu, 2013, “Estimating ground-level PM2.5 concentrations in the southeastern U.S. using geographically weighted regression”, Environmental Research, vol. 212, pp. 1-10.
[29]Z. Kang, Z. Qu, 2017, “Application of BP neural network optimized by genetic simulated annealing algorithm to prediction of air quality index in Lanzhou”, 2017 2nd IEEE International Conference on Computational Intelligence and Applications (ICCIA) (2017), Beijing, China.
[30]Z. Wang, Z. Tian, 2017, “Prediction of Air Quality Index Based on Improved Neural Network”, 2017 International Conference on Computer Systems, Electronics and Control (ICCSEC), Dalian, China.
[31]D. Z. Antanasijević, V. V. Pocajt, D. S. Povrenović, M. Đ. Ristić, A. A. Perić-Grujić, 2013, “PM10 emission forecasting using artificial neural networks and genetic algorithm input variable optimization”, Science of The Total Environment, vol. 443, pp. 511-519.
[32]D. Mishra, P. Goyal, 2016, “Neuro-fuzzy approach to forecast NO2 pollutants addressed to air quality dispersion model over Delhi, India”, Aerosol and Air Quality Research, vol. 16, pp. 166-174.
[33]A. Prakash, U. Kumar, K. Kumar, V. K. Jain, 2011, “A Wavelet-based Neural Network Model to Predict Ambient Air Pollutants’ Concentration”, Environmental Modeling & Assessment, vol. 16, pp. 503-517.
[34]B. T. Ong, K. Sugiura, K. Zettsu, 2016, “Dynamically pre-trained deep recurrent neural networks using environmental monitoring data for predicting PM2.5”, Neural Computing and Applications, vol. 27, pp. 1553-1566.
[35]S. Hochreiter, J. Schmidhuber, 1997, “Long short-term memory”, Neural Comput, vol. 9, pp.1735-1780.
[36]X. Luo, D. Li, Y. Yang, S. Zhang, 2019, “Spatiotemporal Traffic Flow Prediction with KNN and LSTM”, Journal of Advanced Transportation, vol. 2019, pp. 1-10.
[37]M. Abdel-Nasser, K. Mahmoud, 2019, “Accurate photovoltaic power forecasting models using deep LSTM-RNN”, Neural Computing and Applications, vol. 31, pp.2727-2740.
[38]Y. K. Ha, H. W. Chang, 2018, “Forecasting the volatility of stock price index: A hybrid model integrating LSTM with multiple GARCH-type models”, Expert Systems with Applications, vol. 103, pp. 25-37.
[39]D. P. Kingma, J. Ba, 2015, “Adam: A Method for Stochastic Optimization”, ICLR.
[40]D. E. Rumelhart, G. E. Hinton, R. J. Williams, 1986, “Learning representations by back-propagating errors”, Nature, vol. 323, pp. 533-536.
[41]L. A. Zadeh, 1965, “Fuzzy Sets”, Information and Control, vol. 8, pp. 338-353.
[42]L. A. Zadeh, 1973, “Outline of a New Approach to the Analysis of Complex Systems and Decision Processes”, Man and Cybernetics, vol. SMC-3, pp. 28-44.
[43]MathWorks Support Documentation. Available at:
https://www.mathworks.com/help/fuzzy/types-of-fuzzy-inference-systems.html#bu_nr46-1. (Accessed on 30 June 2020)
[44]X. Li, L. Peng, X. Yao, S. Cui, Y. Hu, C. You, T. Chi, 2017, “Long short-term memory neural network for air pollutant concentration predictions: Method development and evaluation”, Environmental Pollution, vol. 231, pp. 997-1004.
[45]J. Ma, J. C.P. Cheng, C. Lin, Y. Tan, J. Zhang, 2019, “Improving air quality prediction accuracy at larger temporal resolutions using deep learning and transfer learning techniques”, Atmospheric Environment, vol. 214, Article 11685.
[46]MathWorks Support Documentation. Available at:
https://www.mathworks.com/help/fuzzy/foundations-of-fuzzy-logic.html. (Accessed on 30 June 2020)
[47]Y. Bai, B. Zeng, C. Li, J. Zhang, 2019, “An ensemble long short-term memory neural network for hourly PM2.5 concentration forecasting”, Chemosphere, vol. 222, pp. 286-294.
[48]Ministry of Justice Investigation Bureau. Available at:
https://www.mjib.gov.tw/FileUploads/eBooks/8b78ec4d0d2147d5b2a71fb976cef156/Section_file/1c076a3a302048a8bdd2ea2938a6b3df.pdf. (Accessed on 30 June 2020)
[49]H.M. Rogers, J.C. Ditto, D.R. Gentner, 2020, “Evidence for impacts on surface-level air quality in the northeastern US from long-distance transport of smoke from North American fires during the Long Island Sound Tropospheric Ozone Study (LISTOS) 2018”, Atmos. Chem. Phys., 20(2020), pp. 671-682.