Derin Öğrenme Modelleri Kullanılarak Emisyon Tahmininin Karşılaştırmalı Çalışması

Öz

Bu çalışma, dizel motorun biyodizel-dizel karışımları ve sıkıştırılmış doğal gaz (CNG) ile çalıştırılması durumunda egzoz emisyonlarının (CO, CO₂ ve NOx) derin öğrenme modelleri kullanılarak tahmin edilmesini incelemektedir. Kanola, ayçiçeği ve mısır yağlarından elde edilen biyodizel, konvansiyonel dizel ile, CNG ise 0, 5, 10 ve 15 litre/dakika (lt/dak) debilerinde motora verilmiştir. İki derin öğrenme mimarisi olan Geçitli yineleme birimi ve uzun kısa süreli bellek emisyonları tahmin etmek için kullanılmıştır. Modellerin performansı, R², RMSE ve Kling-Gupta Efficiency (KGE) metrikleri kullanılarak değerlendirilmiştir. Sonuçlar, her iki modelin de tüm emisyon türleri için R² değerlerinin 0.95'in üzerinde olduğunu ve yüksek doğruluk sağladığını göstermiştir. GRU modeli, CO ve NOx emisyonlarını tahmin etmede daha üstün performans gösterirken, LSTM modeli CO₂ emisyonlarını tahmin etmede daha başarılı olmuştur. Bu çalışma, derin öğrenme modellerinin egzoz emisyonlarını doğru bir şekilde tahmin etme ve çevresel etkiyi azaltmak için yakıt karışımlarını optimize etme potansiyelini vurgulamaktadır.

Anahtar Kelimeler

• Optimizasyon
• Derin Öğrenme
• CNG
• Biyodizel

Comparative Study of Emission Prediction Using Deep Learning Models

Öz

This study investigates the prediction of exhaust emissions (CO, CO₂, and NOx) from a diesel engine fueled with biodiesel-diesel blends and compressed natural gas (CNG) using deep learning models. Biodiesel derived from canola, sunflower, and corn oils was blended with conventional, while CNG was introduced at flow rates of 0, 5, 10, and 15 liters per minute (lt/min). Two deep learning architectures, Gated Recurrent Units (GRU) and Long Short-Term Memory (LSTM), were employed to predict emissions. The models' performance was evaluated using metrics such as R², RMSE, and Kling-Gupta Efficiency (KGE). The results demonstrated that both models achieved high accuracy, with R² and KGE values exceeding 0.93 for all emission types. The GRU model showed superior performance in predicting CO and NOx emissions, while the LSTM model excelled in predicting CO₂ emissions. The study highlights the potential of deep learning models in accurately predicting exhaust emissions and optimizing fuel blends for reduced environmental impact.

Anahtar Kelimeler

• Optimization
• Deep Learning
• CNG
• Biodiesel

Kaynakça

1. 1. Elgohary, M.M., Seddiek, I.S. & Salem, A.M. (2015). Overview of alternative fuels with emphasis on the potential of liquefied natural gas as future marine fuel. Proceedings of the Institution of Mechanical Engineers, Part M: Journal of Engineering for the Maritime Environment, 229(4), 365-375.
2. 2. Zheng, F., Zhang, H., Yin, H., Fu, M., Jiang, H., Li, J. & Ding, Y. (2022). Evaluation of real-world emissions of China V heavy-duty vehicles fueled by diesel, CNG and LNG on various road types. Chemosphere, 303, 135137.
3. 3. Kumari, S. & Singh, S.K. (2023). Machine learning-based time series models for effective CO2 emission prediction in India. Environmental Science and Pollution Research, 30(55), 116601-116616.
4. 4. Pathak, S.K., Nayyar, A. & Goel, V. (2021). Optimization of EGR effects on performance and emission parameters of a dual fuel (Diesel+ CNG) CI engine: an experimental investigation. Fuel, 291, 120183.
5. 5. Sahoo, B.B., Jha, R., Singh, A. & Kumar, D. (2019). Long short-term memory (LSTM) recurrent neural network for low-flow hydrological time series forecasting. Acta Geophysica, 67(5), 1471-1481.
6. 6. Uluocak, I. & Bilgili, M. (2024). Daily air temperature forecasting using LSTM-CNN and GRU-CNN models. Acta Geophysica, 72(3), 2107-2126.
7. 7. Ramachandran, E., Krishnaiah, R., Venkatesan, E.P., Parida, S., Dwarshala, S.K.R., Khan, S.A., ... & Linul, E. (2023). Prediction of RCCI combustion fueled with CNG and algal biodiesel to sustain efficient diesel engines using machine learning techniques. Case Studies in Thermal Engineering, 51, 103630.
8. 8. Sahoo, S., Kumar, V.N.S.P. & Srivastava, D.K. (2022). Quantitative analysis of engine parameters of a variable compression ratio CNG engine using machine learning. Fuel, 311, 122587.

9. 9. Niknam, Y. (2024). NOX emission prediction of a dual-fuel (Diesel+ CNG) compression ignition engine using the DCNN model. Power System Technology, 48(1), 1034-1053.
10. 10. Uludamar, E. (2016). Vibration measurement based analysses of internl Combustion Engines. Doktora Tezi. Çukurova Üniversitesi, Fen Bilimleri Enstitüsü, Makine Mühendisliği Anabilim Dalı, Adana.
11. 11. Bilgili, M., Durhasan, T. & Pinar, E. (2024). Time series analysis of sea surface temperature change in the coastal seas of Türkiye. Journal of Atmospheric and Solar-Terrestrial Physics, 263, 106339.
12. 12. Korkmaz, C. ve Kacar, İ. (2024). Zaman serisinin kestirimi için uzun-kısa süreli bellek ağı yaklaşımı. Çukurova Üniversitesi Mühendislik Fakültesi Dergisi, 39(4), 1053-1066.
13. 13. Liu, Y., Zhang, H., Wu, C., Shao, M., Zhou, L. & Fu, W. (2024). A short-term wind speed forecasting framework coupling a maximum information coefficient, complete ensemble empirical mode decomposition with adaptive noise, shared weight gated memory network with improved northern goshawk optimization for numerical weather prediction correction. Sustainability, 16(16), 6782.
14. 14. Altıparmak, Z. & Aksu, İ.Ö. (2024). Time series installed capacity forecasting with deep learning approach for Türkiye. Cukurova University, Journal of the Faculty of Engineering, 39(3), 709-718.
15. 15. Bilgili, M., Pinar, E. & Durhasan, T. (2025). Global monthly sea surface temperature forecasting using the SARIMA, LSTM, and GRU models. Earth Science Informatics, 18(1), 1-17.
16. 16. Chang, V., Xu, Q.A., Chidozie, A., Wang, H. & Marino, S. (2024). Predicting economic trends and stock market prices with deep learning and advanced machine learning techniques. Electronics, 13(17), 3396.
17. 17. Araya, D., Mendoza, P.A., Muñoz-Castro, E. & McPhee, J. (2023). Towards robust seasonal streamflow forecasts in mountainous catchments: impact of calibration metric selection in hydrological modeling. Hydrology and Earth System Sciences, 27(24), 4385-4408.
18. 18. Bilgili, M. & Pinar, E. (2023). Gross electricity consumption forecasting using LSTM and SARIMA approaches: A case study of Türkiye. Energy, 284, 128575.