Research on the Prediction Model of Cracking Furnace Outlet Temperature Based on CNN-LSTM-Attention

The cracking temperature of the ethylene cracking furnace affects the ethylene yield and has a significant impact on the economic benefits of the cracking furnace. At the same time, the cracking furnace temperature has a great influence on the safe operation of the cracking furnace. Based on the operational data of the cracking furnace process, this article proposes a CNN-LSTM Attention based prediction model for the outlet temperature of an ethylene cracking furnace. Firstly, dynamic features of the data were extracted using convolutional neural networks (CNN). Then, a long short-term memory (LSTM) network was used to learn and train the extracted features. Finally, an attention mechanism was used to allocate different weights to the LSTM hidden layer, predict the outlet temperature of the ethylene cracking furnace, and compare the actual data. The results showed that the average absolute error (MAPE), root mean square error (RMSE), and coefficient of determination (R2) of the CNN-LSTM Attention model were 0.0044, 4.8895, and 0.9846, respectively, Meet the prediction accuracy requirements, with higher prediction accuracy compared to LSTM models. This further proves the rationality of the model prediction and can provide reference for the outlet temperature of the ethylene cracking furnace.