Pet coke gasification via novel gasifier design: Experimental and machine learning based modeling approaches

Creative Commons License

Kiriş B., Yasar M.

INTERNATIONAL JOURNAL OF HYDROGEN ENERGY, cilt.140, ss.452-463, 2025 (SCI-Expanded, Scopus)

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 140
  • Basım Tarihi: 2025
  • Doi Numarası: 10.1016/j.ijhydene.2025.05.365
  • Dergi Adı: INTERNATIONAL JOURNAL OF HYDROGEN ENERGY
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Artic & Antarctic Regions, Chemical Abstracts Core, Chimica, Compendex, Environment Index, INSPEC
  • Sayfa Sayıları: ss.452-463
  • İstanbul Üniversitesi-Cerrahpaşa Adresli: Evet

Özet

Petroleum coke or pet coke is a by-product of petroleum refinery and obtained as a residue from coking process. As a fuel with high carbon content and heating value and low ash content pet coke has a potential to be considered as an alternative and significant energy source. Carbon based fuels such as pet coke, are chemically converted to syngas consisting mainly of a mixture of hydrogen and carbon monoxide by gasification process. Syngas is a versatile raw material used in the production of various chemicals such as ammonia, methanol and many hydrocarbons including synthetic fuels obtained by the Fischer-Tropsch reaction. In this paper, unlike the systems mentioned in literature, a pilot-scale bifunctional gasification system was designed and constructed for gasification of pet coke. The gasification system integrates an entrained flow bed and a cyclone bed. The effects of reaction temperature were investigated in the range of 1123-1323 K. The effects of pet coke feed rate and particle size were also determined. All experimental data were then analyzed using various machine learning algorithms, including Random Forest, M5P, Random Tree, Artificial Neural Network, and Support Vector Machines. The performance of these algorithms was comparatively evaluated. Among the tested models, the Artificial Neural Network demonstrated the highest predictive accuracy, with a correlation coefficient of R = 0.9965 and low prediction errors for H2 yield, including MAE of 1.2715, MAPE of 6.1, RMSE of 2.0701, RAE of 5.20 %, and RRSE of 7.72 %. These findings indicate that machine learning algorithms can serve as effective tools for the operational optimization of pet coke gasification systems.