隨著工業需求增加，廢棄物處理的問題也日益嚴重。目前主要處理廢棄物的方式是依靠焚化爐焚燒。然而，在焚燒過程中，焚化爐會從煙囪排放空氣汙染物導致附近的空氣品質下降。常見的空氣污染物包含了氮氧化物、硫氧化物、粒狀污染物、以及揮發性有機物等，這些汙染對人體的危害影響非常大，如何精確預測空氣汙染物對於環境的維護與人類的健康至關重要。因此本研究使用四種深度學神經網路模型來預測焚化爐的污染排放，包括(1)遞歸神經網路(RNN)、(2)長短期記憶網路(LSTM)、(3) 門控循環單元網路(GRU)以及近年熱門的(4)時間融合變換器網路(TFT)模型等。在資料預處理部分，經由異常值處理將原始數據中的空值與極值做一個合理的轉換。接著確認要預測的目標變量，並使用斯皮爾曼相關係數找出連續性變數之間的關聯性，從而減少模型輸入的複雜度。最後使用Min-max標準化將數據集進行縮放以便增加模型的訓練速度。此外，在模型的參數設定上。本研究針對遞歸模型使用了單層、雙層與三層的神經網路隱藏層，並建立窗口長度大小為5、10、20與30分鐘下預測焚化爐所排放的CO、NOx等空氣汙染物。最終使用平均絕對誤差（MAE）與均方根誤差（RMSE）當評估指標比較模型的預測結果。結果顯示在遞歸模型中，單層與雙層GRU模型的預測效果較佳且訓練時間最短。而當隱藏層數越大時則預測誤差越大。此外，TFT模型是基於注意力架構的DNN模型，對於週期性的數據預測效果良好且擁有時間序列的可解釋性。

This research uses four deep neural network models to predict the pollution emissions of incinerators, including recurrent neural network, long short-term memory network, gated recurrent unit network and the popular Temporal Fusion Transformers network in recent years, etc., In the data preprocessing part, the null values and extreme values in the original data are converted reasonably through outlier processing. Next, identify the target variable to predict and use Spearman's correlation coefficient to find correlations between continuous variables, thereby reducing the complexity of the model input. Finally, the dataset is scaled using Min-max normalization to increase the training speed of the model. In addition, in the parameter setting of the model. This study uses single-layer, double-layer, and three-layer neural network hidden layers for the recursive model, and establishes a window length of 5, 10, 20 and 30 minutes to predict the CO, NOx and other air pollutants emitted by incinerators. Finally, the mean absolute error and the root mean square error are used as evaluation indicators to compare the prediction results of the models. The results show that in the recursive model, the single-layer and double-layer GRU models have better prediction performance and the shortest training time. When the number of hidden layers is larger, the prediction error is larger. In addition, in the TFT model, the prediction effect is good for periodic data, and it has the interpretability of time series.

[1] R. Aiswarya, A. William (2017), Air Quality Impacts of Petroleum Refining and Petrochemical Industries, Environments 4(3), 66-82.
[2] P.W. Tait, J. Brew, A. Che, A. Costanzo (2020), The Health Impacts of Waste Incineration: A Systematic Review, Public Health 44(1), 40-48.
[3] N.P. Cheremisinoff (2003), Volume Reduction Technologies, Handbook of Solid Waste Management and Waste Minimization Technologies, 30-173.
[4] 台灣空氣污染防制法 : https://law.moj.gov.tw/LawClass/LawAll.aspx?pcode=O0020001, access on: August 12, 2022.
[5] E. Pathmanathan, R. Ibrahim, V.S. Asirvadam (2012), Development of CO2 Emission Model of an Acid Gas Incinerator Using Nelder–Mead Least Squares Support Vector Regression, Transactions of the Institute of Measurement and Control 34(8), 974-982.
[6] L. Zheng, S. Yu, M. Yu, Monitoring NOx Emissions from Coal Fired Boilers Using Generalized Regression Neural Network, IEEE paper, 2th International Conference on Bioinformatics and Biomedical Engineering, Shanghai, China, May 16th, 2008.
[7] S.R. Navghare, Dr.G.L. Bodhe (2011), Design of Adaptive Ph controller Using ANFIS, International Journal of Computer Applications 33(6), 41–48.
[8] I. Norhayati, M. Rashidl (2018), Adaptive Neuro-Fuzzy Prediction of Carbon Monoxide Emission from a Clinical Waste Incineration Plant, Neural Computer & Application 30, 3049-3061.
[9] A.L. AhmadaI, A. Azidb, A.R. Yusofb (2004), Emission Control in Palm Oil Mills Using Artificial Neural Network and Genetic Algorithm, Computers & Chemical Engineering 28(12), 2709-2715.
[10] B. Lindemann, T. Müller, H. Vietz, N. Jazdi, M. Weyrich (2021), A Survey on Long Short-Term Memory Networks for Time Series Prediction, Procedia CIRP 99, 650-655.
[11] Y. Lv, J. Liu, T. Yang, D. Zeng (2013), A Novel Least Squares Support Vector Machine Ensemble Model for NOx Emission Prediction of a Coal-Fired Boiler, Energy 55, 319-329.
[12] T. Santos, R. Kern (2016), A Literature Survey of Early Time Series Classification and Deep Learning, Computer Science, 1-7.
[13] H. Shih1, S. Rajendran (2019), Comparison of Time Series Methods and Machine Learning Algorithms for Forecasting Taiwan Blood Services Foundation’s Blood Supply, Journal of Healthcare Engineering, 1-6.
[14] J.F. Torres, D. Hadjout, A. Sebaa (2021), Deep Learning for Time Series Forecasting: A Survey, Big Data 9(1), 3-21.
[15] 台灣固定污染源空氣污染物排放標準 : https://oaout.epa.gov.tw/law/LawContent.aspx?id=FL015350, access on: August 12, 2022.
[16] 台灣廢棄物焚化爐空氣污染物排放標準 ： https://law.moj.gov.tw/LawClass/LawAll.aspx?pcode=O0020010, access on: August 12, 2022.
[17] 中國危險廢物焚燒污染控制標準 : https://www.mee.gov.cn/ywgz/fgbz/bz/bzwb/gthw/gtfwwrkzbz/202012/W020201218699412566946.pdf, access on: August 12, 2022.
[18] Environmental Protection Agency (2016), Federal Plan to Implement Emission Guidelines for Existing CISWI, Federal Register 82(7), 3554-3599.
[19] Japan Environment Governing Standards : https://www.usfj.mil/Portals/80/Documents/2020%20JEGS.pdf?ver=mnuvzSujxiVlhHGcY_m65g%3D%3D, access on: August 12th, 2022.
[20] W.R Niessen (2010). Combustion and Incineration Processes: Applications in Environmental Engineering, Fourth Edition (4th ed.). CRC Press. https://doi.org/10.1201/EBK1439805039.
[21] United States National Academies Press, Waste Incineration & Public Health, Washington, 2000.
[22] A. Kumar, S.R. Samadder (2020), Performance Evaluation of Anaerobic Digestion Technology for Energy Recovery from Organic Fraction of Municipal Solid Waste: A Review, Energy 197(15), 1-13.
[23] M. Kannangara, R. Dua, L. Ahmadi (2018), Modeling and Prediction of Regional Municipal Solid Waste Generation and Diversion in Canada Using Machine Learning Approaches, Waste Management 74, 3-15.
[24] M. Togaçar, B. Ergen, Z. Cömert (2020), Waste Classification Using AutoEncoder Network with Integrated Feature Selection Method in Convolutional Neural Network Models, Measurement 153, 1-10.
[25] S. Kujawa, J. Mazurkiewicz, W. Czekała (2020), Using Convolutional Neural Networks to Classify the Maturity of Compost Based on Sewage Sludge and Rapeseed Straw, Journal of Cleaner Production 258, 1-10.
[26] L. Wang, F. Long, W. Liao, H. Liu (2020), Prediction of Anaerobic Digestion Performance and Identification of Critical Operational Parameters Using Machine Learning Algorithms, Bioresource Technology 298, 1-7.
[27] J. Taoa, R. Lianga, J. Lia, B. Yan (2020), Fast Characterization of Biomass and Waste by Infrared Spectra and Machine Learning Models, Journal of Hazardous Materials 387, 1-9.
[28] H. Cao, Y. Xin, Q. Yuan (2016), Prediction of Biochar Yield from Cattle Manure Pyrolysis Via Least Squares Support Vector Machine Intelligent Approach, Bioresource Technology 202, 158-164.
[29] J. Tang, J. Qiao, Z. Guo, A. Yan, Dioxin Emission Concentration Soft Measuring Method Based on Selective Ensemble Least Square Support Vector Machine Algorithm, CCC paper, 37th Chinese Control Conference, Wuhan, China, July 25th, 2018.
[30] I. Norhayati, M. Rashidl (2018), Adaptive Neuro-Fuzzy Prediction of Carbon Monoxide Emission from a Clinical Waste Incineration Plant, Neural Computer & Application 30, 3049-3061.
[31] Y.M. Ali, Y. Ozgun (2018), An Artificial Intelligence Based Approach to Predicting Syngas Composition for Downdraft Biomass Gasification, Energy 165, 895-901.
[32] O. Yucel, E.S. Aydin (2019), Comparison of the Different Artificial Neural Networks in Prediction of Biomass Gasification Products, International Journal of Energy Research 43(11), 5992-6003.
[33] D. Niu, F. Wu, S. Dai, S. He, B. Wu (2021), Detection of Long-Term Effect in Forecasting Municipal Solid Waste Using a Long Short-Term Memory Neural Network, Journal of Cleaner Production 290, 1-8.
[34] S. Kujawa, J. Mazurkiewicz, W. Czekała (2020), Using Convolutional Neural Networks to Classify the Maturity of Compost Based on Sewage Sludge and Rapeseed Straw, Journal of Cleaner Production 258, 1-10.
[35] M. McCormick, A.E.P. Villa, LSTM and 1-D Convolutional Neural Networks for Predictive Monitoring of the Anaerobic Digestion Process, ICANN paper, 28th International Conference on Artificial Neural Networks, Munich, Germany, September 17th, 2019.
[36] H. You, Z. Ma, Y. Tang, Y. Wang (2017), Comparison of ANN (MLP), ANFIS, SVM, and RF models for the online classification of heating value of burning municipal solid waste in circulating fluidized bed incinerators, Waste Management 68, 186-197.
[37] X. Zhang, Y. Zou, S. Li, S. Xu (2019), A Weighted Auto Regressive LSTM Based Approach for Chemical Processes Modeling, Neurocomputing 367, 64-74.
[38] S. Bunsan, W.Y. Chen (2013), Modeling the Dioxin Emission of a Municipal Solid Waste Incinerator Using Neural Networks, Chemosphere 92(3), 258-264.
[39] C.D. Jayaweera, N. Aziz (2018), Reliability of Principal Component Analysis and Pearson Correlation Coefficient, for Application in Artificial Neural Network Model Development, for Water Treatment Plants, Materials Science and Engineering 458, 1-7.
[40] M. Gevrey, L. Dimopoulos (2003), Review and Comparison of Methods to Study the Contribution of Variables in Artificial Neural Network Models, Ecological Modelling 160(3), 249-264.
[41] K. Wensi, H. Dexian, Soft Sensor Development and Applications Based on LSTM in Deep Neural Networks, IEEE paper, 2017 IEEE Symposium Series on Computational Intelligence, Honolulu, HI, USA, November 27th, 2017.
[42] P.S. Aditya, S.S. Mohan (2020), Performance Prediction and Interpretation of a Refuse Plastic Fuel Fired Boiler, IEEE Access 8, 117467–117482.
[43] S. Hochreiter (1998), The Vanishing Gradient Problem During Learning Recurrent Neural Nets and Problem Solutions, Fuzziness and Knowledge-Based Systems 6(2), 107-116.
[44] K. Myroslav, Y. Pavlo, Deep Neural Network for Image Recognition Based on the Caffe Framework, IEEE paper, 2018 IEEE Second International Conference on Data Stream Mining & Processing, August 21th, 2018.
[45] M. Andrew, Q. Peng (2017), Building DNN Acoustic Models for Large Vocabulary Speech Recognition, Computation Speech and Language 41, 195-213.
[46] T. Pál, G. Attila (2017), Image-Based Deep Neural Network Prediction of the Heat Output of a Step-Grate Biomass Boiler, Applied Energy 200, 155-169.
[47] Ö. Yngve, T. Pál (2018), Development of a Vision-Based Soft Sensor for Estimating Equivalence Ratio and Major Species Concentration in Entrained Flow Biomass Gasification Reactors, Applied Energy 226(15), 450-460.
[48] T. Kanekol, Y. Tsuruminel, J. Poon, Learning Deep Dynamical Models of a Waste Incineration Plant from In-Furnace Images and Process Data, IEEE paper, 15th International Conference on Automation Science and Engineering, Vancouver, BC, Canada, September 19th, 2019.
[49] C.L. Zachary, B. John (2015), A Critical Review of Recurrent Neural Networks for Sequence Learning, Computer Science 4, 1-38.
[50] S. Hochreiter (1998), The Vanishing Gradient Problem During Learning Recurrent Neural Nets and Problem Solutions, Fuzziness and Knowledge-Based Systems 6(2), 107-116.
[51] S. Hochreiter, J. Schmidhuber (1997), Long Short-Term Memory, Neural Computation 9(8), 1735-1780.
[52] J. Chung, C. Gulcehre, K. Cho, Y. Bengio (2014), Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modeling, Neural and Evolutionary Computing, 1-9.
[53] D. Rahul, M.S. Fathi, Gate-Variants of Gated Recurrent Unit (GRU) Neural Networks, IEEE paper, 60th International Midwest Symposium on Circuits and Systems, Boston, MA, USA, August 6th, 2017.
[54] B. Lim, S.O. Arık, N. Loeff, T. Pfister (2019), Temporal Fusion Transformers for Interpretable Multi-horizon Time Series Forecasting, Machine Learning, 1-27.