Waste to energy: Enhancing biogas utilization in dual-fuel engines using machine learning based prognostic analysis

Paramasivam P., Alruqi M., AĞBULUT Ü.

Fuel, cilt.381, 2025 (SCI-Expanded) identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 381
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1016/j.fuel.2024.133093
  • Dergi Adı: Fuel
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, PASCAL, Aerospace Database, Biotechnology Research Abstracts, Chemical Abstracts Core, Chimica, Communication Abstracts, Compendex, INSPEC, Metadex, Pollution Abstracts, Civil Engineering Abstracts
  • Anahtar Kelimeler: Alternatve fuels, Emisison, Explainable machine learning, Predictive modeling, Violin plots, XGBoost
  • Yıldız Teknik Üniversitesi Adresli: Hayır

Özet

Alternative fuels derived from organic matter like biomass can be a viable solution in the present scenario of increasing greenhouse gases. In the present study, waste food and animal refuse-derived biogas were tested as fuel in a diesel engine. To further enhance the waste-to-energy perspective waste cooking oil biodiesel − diesel blends were used as pilot fuel. A comprehensive set of operational settings including fuel injection timing, pressure, compression ratios, different flow rates of biogas, and engine loads were tested. Four machine learning approaches Linear Regression (LR), Decision Tree (DT), Extreme Gradient Boosting (XGBoost), and Gaussian Process Regression (GPR) were employed to develop prognostic models. The developed models for engine performance and emission were predicted with high accuracy with R2 values as high as 0.9962 for brake-specific fuel consumption and 0.9948 for brake thermal efficiency. The prediction errors were low 0.06 for BTE during model training and 0.18 during the model testing phase while it was almost negligible in the case of BSFC. All models were compared using statistical metrics and violin plots. The DT-based forecasting models were observed to be the best among all the models both based on statistical measures as well as violin plots.