Nowadays

circulating fluidized bed(CFB) coal-fired boilers face challenges in the process of deep peak regulation

such as high CO emission concentrations and the lack of theoretical guidance for collaborative emission reduction of multiple pollutants including NO_x and SO_2. Taking a 150 t/h CFB coal-fired boiler as the research object

a model for quickly predicting mass concentrations of CO

NO_x and SO_2 emitted from the furnace is established based on the long short-term memory(LSTM) neural network

the Attention mechanism and the XGBoost algorithm. Moreover

an online emission reduction strategy is proposed by coupling with the particle swarm optimization(PSO) algorithm. 36 298 operational data points from the coal-fired boiler throughout 2023 are selected as training samples. A correlation analysis is conducted between the boiler inspection data and pollutant emission mass concentrations to determine the input parameters for the prediction model. The fitness function and boundary function are determined with the prediction model coupled with the PSO algorithm. Through the calculation of emission reduction optimization model

an online emission reduction optimization strategy for CO

NO_x and SO_2 mass concentrations of CFB boilers in different load ranges is proposed

and the feasibility of the algorithm in practical boiler tuning applications is evaluated.