Temperature Modeling Of Boiler Combustion Layers Based On Data Analysis And Mining

Furnace combustion directly affects the safety,economy and environmental protection of boiler operation.Furnace combustion layer temperature not only directly reflects the quality of coal into the furnace,but also is a key factor in the formation of combustion pollutants.Based on data mining and data analysis,aiming at the boiler operation data of 660 MW ultra-supercritical coal-fired units,the static and dynamic models of combustion layers' temperature are established by VIP?SR regression method,NARX neural network,ARDL model and ARIMAX model,respectively.The main research contents are as follows:Firstly,the main algorithm characteristics of data mining and data analysis are analyzed,including partial least squares regression,stepwise regression,ARDL model and ARIMAX model.Aiming at the boiler combustion system of 660 MW ultra supercritical unit,the operation data under low,medium and high loads are collected,the main parameters affecting the temperature of each combustion layer under different loads are analyzed,and the variables are screened based on the VIP technology.The main changes affecting the temperature of combustion layers in furnace are obtained from the correlation among the parameters.The number of independent variables in the sample set is reduced,and the collinearity between variables is eliminated,which lays a foundation for combustion layer temperature modeling.Secondly,the variables are screened by the variable importance for projection of PLS,and the variables are further simplified by stepwise regression method.Based on VIP-SR,the temperature prediction model of furnace combustion layers under different loads are established,and the accuracy of the model is tested on the test set.At the same time,the SVM method is used to complete the temperature prediction of each combustion layer under low load,and the characteristics of the two temperature prediction models are compared and analyzed.Finally,based on the non-linear active autoregressive neural network(NARX),the autoregressive distributed lag(ARDL)model and the active autoregressive integral moving average model(ARIMAX),the dynamic modeling of the three-layer combustion layer temperature in the furnace under different loads is realized.The simulation results show the effectiveness of the algorithm and the prediction of different temperatures.The model was compared and analyzed.