A new online optimization method for boiler combustion system based on the data-driven technique and the case-based reasoning principle

To adapt to the time-variability of boiler combustion systems, a new online combustion optimization method for boiler is proposed in this paper. The massive historical combustion data are preprocessed, and then an improved constrained fuzzy weighted rule is employed to extract combustion rules from historical combustion data. After that, an improved particle swarm optimization-based least square support vector machine is adopted to construct the dynamic mathematic model for boiler efficiency and NOx emission, respectively, and an improved multi-objective particle swarm optimization algorithm based on the well-construction dynamic mathematical model is proposed and applied to excavate deeply the combustion rules of boiler, and the optimization case library is constructed by integrating all combustion rules. At last similarity measure-based case-based reasoning method is employed to rapidly identify the well-performance similar cases from the optimization case library, which is helpful to complete the online combustion optimization. The effectiveness of proposed online optimization method for boiler is proved by applying it to an actual combustion process. The results showed that proposed online optimization method could take less time to gain a set of excellent operating solution, the NOx emission reduced by 9.236% on average and the boiler efficiency increased by 0.046% on average. Therefore, the pro-posed online combustion optimization method for boiler has the ability to realize the online combustion opti-mization of boiler.