Abstract: As an important energy conversion equipment,boiler plays an important role in China’s industrial field.By 2020,the pollutant emission and energy consumption of industrial boilers in China are second only to power station boilers.Industrial boilers have the following characteristics:（1）they are mainly coal-fired,accounting for more than 80% of the total industrial boilers in use.（2）the pollution problem is serious.The annual emission of nitrogen oxides from the industrial boiler industry was 2-3 million tons,accounting for 15% of the total emission of nitrogen oxides in China.（3）the operation management level is low.Under the same boiler operation conditions,the difference of operator can lead to the boiler operation efficiency difference from 3% to 10%.Therefore,it has become an unavoidable research topic to optimize the operation parameters of industrial boilers and provide suggestions for stokers to improve the operation efficiency of industrial boilers and reduce the emission of pollutants.With the development of artificial intelligence technology,data-driven boiler operation optimization has become a research hotspot.At present,there are still some defects in these studies: using single algorithm such as ANN and LR to model the boiler combustion system;The optimization objective always was single,which was difficult to meet the dual optimization requirements of environmental protection and economy.In view of these problems,this paper carried out the following researchThe research status of combustion modeling and operation optimization of coal-fired boiler is investigated,and the basic structure and operation mechanism of coal-fired boiler are introduced.Based on economy and environmental protection,two optimization objectives of boiler thermal efficiency and NOx emission concentration are determined,and the factors influencing boiler thermal efficiency and NOx emission concentration are analyzed.Aiming at the problem of combustion modeling of coal-fired boiler,a combustion model based on stacking ensemble learning is proposed.The model takes boiler operating parameters and operating parameters as input,boiler thermal efficiency and NOx emission concentration as output.Firstly,linear regression（LR）,support vector regression（SVR）,neural network algorithm（ANN）and extreme gradient lifting tree algorithm（xgboost）are used to establish four basic models,and particle swarm optimization（PSO）is used to optimize the super parameters of ANN and xgboost models to obtain pso-ann and PSO xgboost models.Then,LR,SVR,pso-ann and PSO xgboost models are integrated by stacking ensemble learning algorithm to establish the final combustion model.Experiments show that the accuracy of the integrated combustion model based on stacking algorithm is higher than that of the single algorithm model.Aiming at the operation optimization problem of coal-fired boiler combustion system,based on the established ensemble combustion model of coal-fired boiler,a multi-objective optimization model of thermal efficiency and NOx emission of coal-fired boiler is established by using a weighted evaluation function.The optimization proportion of the two indexes can be changed by adjusting the weighted weight.A particle swarm optimization algorithm with differential attenuation weight is used to optimize the boiler operation parameters,which is compared with a variety of intelligent optimization algorithms.The experimental results show that the optimized boiler operation parameters can make the boiler thermal efficiency and NOx emission concentration better than the historical optimal values under the experimental conditions,and the global and local search ability of the differential attenuation weighted particle swarm optimization algorithm is better than other algorithms.According to the requirements of Jiangsu special inspection institute,the industrial boiler operation optimization platform is designed and implemented.The platform is developed with go language,including the server-side implementation of the industrial boiler operation parameter acquisition system.Optimizes the operation parameters according to the collected boiler operation data,and puts forward suggestions to the operator.
