Abstract: Due to the unstable combustion process, local over-temperature may occur on the tube of the heating furnace which is prone to coking, leading to its failure. Therefore, it is necessary to measure the temperature of the heating furnace in the actual project. At present, computational fluid dynamics (CFD) numerical method is used for direct real-time temperature field prediction. Although the method is of high accuracy, it is very time-consuming and the prediction cannot be achieved with the existing calculation capability. To solve this problem, this paper proposes a real-time prediction method for the temperature field of heating furnaces based on a convolutional long short-term memory network. The method is completely driven by data and uses ConvLSTM to extract the internal relationship of the time sequence dynamically for temperature field prediction. This can realize the soft sensor of the temperature field of industrial heating furnace and the MAE (mean absolute error) of the predicted temperature field is 31.7 K.

Key words: ConvLSTM, softsensor, temperature field of heating furnace, coking, furnace