Prediction Of Flue Oxygen Content In Thermal Power Unit Combustion System

At present,energy conservation and environmental protection are the problems that all power generation enterprises must face.Improving boiler operation efficiency is the key to reducing coal consumption of power generation.In the boiler control system,the oxygen content of the flue gas is one of the important parameters reflecting the operating conditions.The accuracy of the measurement results directly affects the safety and economy of the unit.In the actual production process,the on-line detection of oxygen is affected by harsh environmental factors such as high temperature and high pressure,and the accuracy is difficult to guarantee.The proposed soft sensor technology effectively solves this problem.In order to effectively measure the oxygen quantity and reduce the measurement error,this paper uses the improved particle swarm optimization algorithm to support the support vector regression machine(IPSO-SVR)method to establish a nonlinear prediction model of flue gas oxygen content,and simulate and predict the results.Compared.In this paper,the oxygen content of flue gas in supercritical units is taken as the research object.First,by studying the mechanism and correlation analysis of the boiler combustion system,the auxiliary variables needed to construct the prediction model are preliminarily determined.Then,the kernel principal component analysis(KPCA)is used to reduce the dimension of the input variables,and the eight auxiliary variables affecting the oxygen content are selected.Second,in terms of modeling,the improved particle swarm optimization(IPSO)algorithm is used to optimize the SVR kernel parameters.Since static modeling can not meet the oxygen prediction under multi-conditions,the state constraint is introduced to improve the traditional sliding time window algorithm,and the improved algorithm is used to dynamically correct the model.Based on the field data,the simulation results show that the dynamic prediction model can accurately predict the oxygen content of the system without adding additional measuring points,and can meet the high prediction accuracy and robustness of the model.Requirements can provide a basis for further optimization of coal-fired units.