EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

The Prediction Method for Coil Outlet Temperature of Ethylene Cracking Furnace Based on CNN–BiLSTM–Attention

Xiangwu Deng, Zhiping Peng, Delong Cui, Qirui Li, Ziqin Hong

International Journal of Chemical Engineering · 2026-01 · 2026 (1)

DOI: 10.1155/ijce/8850320 ↗

Kurz gefasst

Fachartikel in der Zeitschrift International Journal of Chemical Engineering, online erschienen am 27.04.2026. Verfasst von Xiangwu Deng und 4 weiteren Autor:innen. Die Arbeit nennt 40 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0). Gefördert von: National Natural Science Foundation of China, Basic and Applied Basic Research Foundation of Guangdong Province, Department of Education of Guangdong Province.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Ethylene is an important raw material, and ethylene cracking furnaces (ECFs) are the core equipment sources of ethylene plants. The quality of the operation of the ECF directly affects the overall economic benefits of the entire plant. The coil outlet temperature (COT) of the ECF tube is a key control indicator for the operation of the ECF, which affects not only the judgment of process personnel regarding the operating status of the ECF but also the safe operating cycle of the ECF and the yield of the main product, ethylene. To accurately predict the COT of the ECF, a method that combines convolutional neural networks (CNNs), bidirectional long short‐term memory (BiLSTM) networks, and attention mechanisms (AEs) is proposed. Feature extraction from the operation data of cracking furnaces was based on CNNs, while BiLSTM offers the advantages of effectively handling the long‐term dependencies of these data and better capturing their long‐term trends and periodic changes in the time dimension. This method combines CNN and BiLSTM to capture the complex nonlinear spatiotemporal relationships of the COT and then uses AEs to focus on important input features. Compared with CNN–BiLSTM model, CNN–LSTM–attention model, CNN–LSTM model and LSTM model, the mean absolute error (MAE), root mean square error (RMSE), mean absolute percentage error (MAPE), and coefficient of determination ( R 2 ) of the proposed CNN–BiLSTM–attention prediction model are best, which were 3.4601, 4.3781, 0.0039, and 0.9897, respectively, and the accuracy of the measurement results are higher than those of the other conventional models under the same conditions. The forecast effect of the deep learning model shows that the presented method has the superior predictive performance, which indicates the feasibility and progressiveness. These results can promote the safe and stable operation of ECFs, thereby providing a theoretical basis for improving the ethylene yield of ECFs.

Förderung

  • National Natural Science Foundation of China
  • Basic and Applied Basic Research Foundation of Guangdong Province
  • Department of Education of Guangdong Province
  • Guangdong University of Petrochemical Technology

Zitieren

APA

Deng, X., Peng, Z., Cui, D., Li, Q., & Hong, Z. (2026). The Prediction Method for Coil Outlet Temperature of Ethylene Cracking Furnace Based on CNN–BiLSTM–Attention. International Journal of Chemical Engineering, 2026(1). https://doi.org/10.1155/ijce/8850320

Vancouver

Deng X, Peng Z, Cui D, Li Q, Hong Z. The Prediction Method for Coil Outlet Temperature of Ethylene Cracking Furnace Based on CNN–BiLSTM–Attention. International Journal of Chemical Engineering. 2026;2026(1). doi:10.1155/ijce/8850320

BibTeX
@article{Deng2026,
  author = {Deng, Xiangwu and Peng, Zhiping and Cui, Delong and Li, Qirui and Hong, Ziqin},
  title = {The Prediction Method for Coil Outlet Temperature of Ethylene Cracking Furnace Based on CNN–BiLSTM–Attention},
  journal = {International Journal of Chemical Engineering},
  year = {2026},
  volume = {2026},
  number = {1},
  doi = {10.1155/ijce/8850320}
}
Kostenlos starten →