Collaborative Multi-Objective Optimization of Combustion and Emissions in Circulating Fluidized Bed Boilers Using the Bidirectional Temporal Convolutional Network and Hybrid Dung Beetle Optimizer

With the increasing global focus on sustainable development, circulating fluidized bed (CFB) boilers, as highly efficient and low-pollution combustion equipment, play an important role in energy production and environmental protection. However, the combustion efficiency and emission control of CFB boilers still face challenges, and there is an urgent need for multi-objective optimization through advanced technologies to support the goal of sustainable development. This study proposes an intelligent framework integrating Bidirectional Temporal Convolutional Network (BiTCN) and Hybrid Dung Beetle Optimizer (HDBO) for multi-objective optimization of combustion efficiency and NOx/SO 2 emissions in CFB boilers. The BiTCN model captures bidirectional temporal dependencies between dynamic parameters (e.g., air-coal ratio, bed temperature) and target variables through operational data analysis. Three key improvements are implemented in DBO: (1) Chaotic initialization via sequential pattern mining (SPM) enhances population diversity and spatial coverage; (2) The osprey optimization algorithm (OOA) hunting mechanism replaces the original rolling update strategy, improving global exploration; (3) t-Distribution perturbation is applied to foraging beetles in later iterations, leveraging its “sharp peak and thick tail” characteristics to dynamically balance exploitation and exploration. Experimental results demonstrate 0.5–1% combustion efficiency improvement and 15.1%/30% reductions in NOx/SO 2 emissions for a typical CFB boiler.