Monitoring and Optimization of CFB Bed Temperature in the Flexible Process: A Hybrid Framework of Deep Learning Model...

In response to the issues of abnormal bed temperature fluctuations and inefficient combustion that occur during the flexible operation of the circulating fluidized bed (CFB) boiler combustion system, which is characterized by strong coupling, nonlinearity, and large inertia, this paper presents a novel monitoring and optimization approach that integrates a deep learning model with a mechanism model. Specifically, the Informer algorithm is utilized to construct a bed temperature range prediction model, thus enabling the real-time monitoring of the intricate change trends of the in-furnace bed temperature. Considering the poor reliability of combustion optimization target values derived from data model predictions and the constraints on the optimization upper limit, this article further combines the mechanism model to comprehensively determine the combustion optimization target values from the design perspective. The results indicated that the Informer prediction model exhibited a root-mean-square error (RMSE) of 3.385 °C, a mean absolute error (MAE) of 2.45 °C, and a mean absolute percentage error (MAPE) of 0.268% on the rolling test data set, demonstrating high overall prediction accuracy. The steady-state operational data of the generating unit at 300 and 200 MW were selected to validate the mechanism model, with particular emphasis on comparing the distribution trends of pressure and temperature along the height of the furnace. The outcomes revealed good consistency, indicating the model's high accuracy in performance simulation. By integrating the data model with the mechanism model, the target values for optimizing bed temperature performance under steady-state conditions were determined. In comparison to using the data model alone, during the 240 and 280 MW operating conditions, the average thermal efficiency of the boiler increased by 0.19 and 0.13%, respectively. Concurrently, the coal consumption rate for power generation decreased by 0.6707 and 0.4453 g/(kW·h), respectively, and the carbon emissions reduced per kilowatt-hour of electricity generated were 1.6738 and 1.1113 g, respectively.