隨著能源效率和環境可持續性成為全球關注的重點，優化燃燒過程和提升燃燒系統的效能日益重要。在這樣的背景下，人工智慧（AI）技術被引入燃燒系統中，以自動化處理工作項目、識別潛在問題並創造更快的實驗流程。這不僅提升了結果的品質，還降低了實驗成本和浪費。特別是在爐子的設計和運作中，展示了改進的革命性潛力。
本研究運用人工智慧技術預測能源領域中排氣燃燒爐後端產品，採用多種機器學習模型如神經網絡、隨機森林等，通過優化數據工程方法提升數據處理效率。研究成果顯示，經過數據清理和模型調整後，NOX和SOX的預測準確度達到0.9以上，O2和CO也顯著提升至0.92和0.82。此外，通過調整排氣輸入，顯著提高了蒸汽產量和排氣利用效率，證明了調整操作策略對提升能效和環境合規性的重要性。本研究的成果有助於優化生產過程和資源管理，提供了操作效率提升和能源節約的有效策略。
在能源應用領域中，液滴實驗在研究燃燒過程中扮演著關鍵角色，顯著影響了對燃料的熱反應和氣化行為的理解和優化。利用人工智慧，特別是機器學習和深度學習模型，可以有效預測和分析高溫下的滴狀行為，從而指導實驗設計和優化燃燒策略。
本研究也同時提出利用LSTM和XGBoost模型成功預測液滴微爆行為，透過時間序列數據分析不同溫度和化學成分下的特性。實驗結果顯示，短期預測時模型表現優異，但隨時間延長或在高溫條件下，如400°C與500°C，預測精度有所下降，指出高溫可能引入更多非線性因素，增加預測挑戰，而本研究同步探討液滴實驗非時序性之回歸問題，本研究採用類神經網路以及隨機森林模型在預測燃料類型時展示出穩定性與可靠性。當模型於特定溫度範圍（如300°C至400°C）訓練時，對於更高溫度（如500°C）的預測準確度有顯著下降。然而，當擴展訓練範圍至更廣的溫度（包括500°C）並調整數據分配後，模型的預測效能顯著提升，R²值超過0.9，有效預測各種溫度條件下的燃燒行為，顯示出模型在廣泛溫度範圍內的高度適應性與精確性。
本研究通過擬和排氣燃燒爐之實驗數據通過機器學習技術展示了模擬數據方法的實用性，為生產過程的優化提供了強有力的指導。利用機器學習模型在分析和預測液滴微爆特性方面的有效性，並透過逐步改善預測準確度來提升模型性能。這些進展反映了數據處理和模型優化策略的有效性，增強了模型的預測能力，也為未來在類似條件下設計和操作燃燒系統提供了實證支持和創新思路，有助於推動相關環境政策的制定和能源管理策略，進一步促進環境保護和能源應用的發展目標。

As energy efficiency and environmental sustainability become global priorities, optimizing combustion processes and enhancing the performance of combustion systems has become increasingly important. Against this backdrop, artificial intelligence (AI) technologies have been introduced into combustion systems to automate tasks, identify potential issues, and create faster experimental workflows. This not only enhances the quality of results but also reduces experimental costs and waste. Particularly in the design and operation of furnaces, AI has demonstrated revolutionary potential for improvement.
This study employs AI technologies to predict backend products of exhaust combustion furnaces in the energy sector, utilizing various machine learning models such as neural networks and random forests, and enhances data processing efficiency through optimized data engineering methods. The findings indicate that after data cleaning and model tuning, the prediction accuracy for NOX and SOX reached above 0.9, with significant improvements for O2 and CO to 0.92 and 0.82, respectively. Moreover, by adjusting the combustion input parameters, steam production and exhaust utilization efficiency were significantly enhanced, demonstrating the importance of operational strategy adjustments for improving energy efficiency and environmental compliance. The outcomes of this study aid in optimizing production processes and resource management, providing effective strategies for operational efficiency and energy saving.
In the field of energy applications, droplet experiments play a crucial role in studying combustion processes, significantly influencing the understanding and optimization of fuel's thermal reactions and vaporization behaviors. Using artificial intelligence, especially machine learning and deep learning models, effective predictions and analyses of droplet behaviors at high temperatures can guide experimental design and optimize combustion strategies.
The study also successfully predicted droplet micro-explosion behaviors using LSTM and XGBoost models, analyzing characteristics under different temperatures and chemical compositions through time-series data. Experimental results show that models perform excellently in short-term predictions, but the prediction accuracy declines over time or under high temperatures, such as 400°C and 500°C, indicating that higher temperatures may introduce more nonlinear factors, increasing predictive challenges.
Additionally, the regression issues in droplet experiments are explored through non-time-series analysis. The study uses neural network and random forest models to show stability and reliability in predicting fuel types. When models are trained within a specific temperature range (e.g., 300°C to 400°C), the accuracy of predictions at higher temperatures (e.g., 500°C) significantly drops. However, when the training range is expanded to include higher temperatures and the data allocation is adjusted, model performance significantly improves, with an R² value exceeding 0.9, effectively predicting combustion behaviors under various temperature conditions and showing high adaptability and accuracy across a broad temperature range.
This research demonstrates the practicality of simulated data methods through machine learning technology by fitting experimental data from exhaust combustion furnaces, providing strong guidance for optimizing production processes. The effectiveness of machine learning models in analyzing and predicting droplet micro-explosion characteristics, and improving model performance by enhancing prediction accuracy, reflects the effectiveness of data processing and model optimization strategies. These advancements not only enhance the predictive capability of models but also provide empirical support and innovative ideas for designing and operating combustion systems under similar conditions in the future, aiding in the formulation of related environmental policies and energy management strategies, and further promoting environmental protection and energy application development goals.

[1] McCulloch, W. S., and Pitts, W., "A logical calculus of the ideas immanent in nervous activity," Bulletin of Mathematical Biology, 52(1), pp. 99-115, 1990
[2] "https://aws.amazon.com/tw/what-is/neural-network/,"
[3] Casas, I., " Neural Networks," International Encyclopedia of Human Geography, pp. 381-385, 2020
[4] "https://ithelp.ithome.com.tw/articles/10303151?sc=rss.iron,"
[5] Zhou, L., Song, Y., Ji, W., and Wei, H., "Machine learning for combustion," Energy and AI, 7, p. 100128, 2022
[6] Sherstinsky, A., "Fundamentals of Recurrent Neural Network (RNN) and Long Short-Term Memory (LSTM) network," Physica D: Nonlinear Phenomena, 404, p. 132306, 2020
[7] Li, J., Li, L., Tong, Y. W., and Wang, X., "Understanding and optimizing the gasification of biomass waste with machine learning," Green Chemical Engineering, 4(1), pp. 123-133, 2023
[8] Han, Z., Xie, Y., Moinul Hossain, M., and Xu, C., "A hybrid deep neural network model for NOx emission prediction of heavy oil-fired boiler flames," Fuel, 333, p. 126419, 2023
[9] Ilamathi, P., Selladurai, V., Balamurugan, K., and Sathyanathan, V. T., "ANN–GA approach for predictive modeling and optimization of NOx emission in a tangentially fired boiler," Clean Technologies and Environmental Policy, 15(1), pp. 125-131, 2013
[10] Liukkonen, M., Heikkinen, M., Hiltunen, T., Hälikkä, E., Kuivalainen, R., and Hiltunen, Y., "Artificial neural networks for analysis of process states in fluidized bed combustion," Energy, 36(1), pp. 339-347, 2011
[11] Krzywański, J., and Nowak, W., "Neurocomputing approach for the prediction of NOx emissions from CFBC in air-fired and oxygen-enriched atmospheres," Journal of Power Technologies, 97(2), 2017
[12] Jia, M., Wang, X., Zhang, W., Song, Q., Qian, B., Ye, Y., Xu, K., and Wang, X., "Prediction of CO/NOx emissions and the smoldering characteristic of sewage sludge based on back propagation neural network," Environmental Pollution, 342, p. 123049, 2024
[13] Barboza, A. B. V., Mohan, S., and Dinesha, P., "On reducing the emissions of CO, HC, and NOx from gasoline blended with hydrogen peroxide and ethanol: Optimization study aided with ANN-PSO," Environmental Pollution, 310, p. 119866, 2022
[14] Zhu, Q., Jones, J. M., Williams, A., and Thomas, K. M., "The predictions of coal/char combustion rate using an artificial neural network approach," Fuel, 78(14), pp. 1755-1762, 1999
[15] Su, S., and Wang, J., "Machine learning prediction of contents of oxygenated components in bio-oil using extreme gradient boosting method under different pyrolysis conditions," Bioresource Technology, 379, p. 129040, 2023
[16] Ascher, S., Watson, I., and You, S., "Machine learning methods for modelling the gasification and pyrolysis of biomass and waste," Renewable and Sustainable Energy Reviews, 155, p. 111902, 2022
[17] Kargbo, H. O., Zhang, J., and Phan, A. N., "Optimisation of two-stage biomass gasification for hydrogen production via artificial neural network," Applied Energy, 302, p. 117567, 2021
[18] Zhang, L., Zhu, G., Chao, Y., Chen, L., and Ghanbari, A., "Simultaneous prediction of CO2, CO, and NOx emissions of biodiesel-hydrogen blend combustion in compression ignition engines by supervised machine learning tools," Energy, 282, p. 128972, 2023
[19] Jian An, H. W., Bing Liu, Kai Hong Luo, Fei Qin, Guo Qiang He, "A deep learning framework for hydrogen-fueled turbulent combustion simulation," International Journal of Hydrogen Energy, 45(35), 2020
[20] Licheng Zhang, X. T., Yuhan Zhao, Lulu Liu, Zhiwei Li, Lixin Tao, Xiaonan Wang, Xiuhua Guo, Yanxia Luo, "Application of nonlinear land use regression models for ambient air pollutants and air quality index," Atmospheric Pollution Research, 12(10), 2021
[21] Silvio Cesar de Lima Nogueira, S. H. O., Luis Mauro Moura, Eric Domingues, Leandro dos Santos Coelho, Viviana Cocco Mariani, "Prediction of the NOx and CO2 emissions from an experimental dual fuel engine using optimized random forest combined with feature engineering," Energy, 2023
[22] Min-Yuan Cheng, M.-T. C., "Accurately predicting building energy performance using evolutionary multivariate adaptive regression splines," Applied Soft Computing, 22, 2014
[23] Wu, L., and Noels, L., "Recurrent Neural Networks (RNNs) with dimensionality reduction and break down in computational mechanics; application to multi-scale localization step," Computer Methods in Applied Mechanics and Engineering, 390, p. 114476, 2022
[24] Zhao, F., Ruan, Z., Yue, Z., Hung, D. L. S., Som, S., and Xu, M., "Time-sequenced flow field prediction in an optical spark-ignition direct-injection engine using bidirectional recurrent neural network (bi-RNN) with long short-term memory," Applied Thermal Engineering, 173, p. 115253, 2020
[25] Ma, D., Yao, Q., Liu, Y., Wang, L., Ma, L., Wang, J., Sun, M., and Ma, X., "Prediction of volatile composition of coal pyrolysis under various swelling and catalytic conditions based on recurrent neural network," Journal of Analytical and Applied Pyrolysis, 175, p. 106187, 2023
[26] Faridi, I. K., Tsotsas, E., Heineken, W., Koegler, M., and Kharaghani, A., "Spatio-temporal prediction of temperature in fluidized bed biomass gasifier using dynamic recurrent neural network method," Applied Thermal Engineering, 219, p. 119334, 2023
[27] Yang, G., Wang, Y., and Li, X., "Prediction of the NOx emissions from thermal power plant using long-short term memory neural network," Energy, 192, p. 116597, 2020
[28] Li, T., Zhang, Z., and Chen, H., "Predicting the combustion state of rotary kilns using a Convolutional Recurrent Neural Network," Journal of Process Control, 84, pp. 207-214, 2019
[29] Arsie, I., Marra, D., Pianese, C., and Sorrentino, M., "Real-Time Estimation of Engine NOx Emissions via Recurrent Neural Networks," IFAC Proceedings Volumes, 43(7), pp. 228-233, 2010
[30] AG, I. E., "廢氣燃燒爐控制," 專利編號: CN107420917A, 2016
[31] S. Lek, Y. S. P., "Artificial Neural Networks," pp. 237-245, 2008
[32] Hastie, T. J., "Generalized Additive Models (1st ed.)," 1990
[33] Stavropoulos, G., van Voorstenbosch, R., van Schooten, F.-J., and Smolinska, A., "3.32 - Random Forest and Ensemble Methods," Comprehensive Chemometrics (Second Edition), S. Brown, R. Tauler, and B. Walczak, eds., Elsevier, Oxford, pp. 661-672, 2020
[34] Friedman, J. H., "Multivariate Adaptive Regression Splines," The Annals of Statistics, 19, 1991
[35] Mu, Y., Liu, X., and Wang, L., "A Pearson’s correlation coefficient based decision tree and its parallel implementation," Information Sciences, 435, pp. 40-58, 2018
[36] Puth, M.-T., Neuhäuser, M., and Ruxton, G. D., "Effective use of Pearson's product–moment correlation coefficient," Animal Behaviour, 93, pp. 183-189, 2014
[37] library, P., "[Online Available]: https://pandas.pydata.org/,"
[38] library, N., "[Online Available]: https://numpy.org/,"
[39] library, M., "[Online Available]: https://matplotlib.org/,"
[40] library, S., "[Online Available]: https://seaborn.pydata.org/,"
[41] Tsu-Hsuan Lin, T.-H. L., "“Droplet Lifetime and Micro-Explosion of Fuel Drops Emulsified with Hydrogen Peroxide”," 2022
[42] Chen, X., Weng, T., Li, C., and Yang, H., "Equivalence of machine learning models in modeling chaos," Chaos, Solitons & Fractals, 165, p. 112831, 2022
[43] Zhang, X., Li, H., Sekar, M., Elgendi, M., Krishnamoorthy, N. R., Xia, C., and Priya Matharasi, D., "Machine learning algorithms for a diesel engine fuelled with biodiesel blends and hydrogen using LSTM networks," Fuel, 333, p. 126292, 2023
[44] Sundermeyer, M., Schlüter, R., and Ney, H. Published, "Lstm neural networks for language modeling," Proc. Interspeech, pp. 194-197, 2012