Thermal Efficiency Prediction And Combustion Optimization Of Reheating Furnace Based On Improved DBN

With the increasing demand for energy,there is a situation of high consumption and low utilization in industrial production.The heating furnaces commonly used in the petrochemical industry have the problem of insufficient combustion.Therefore,improving the combustion efficiency of the heating furnaces is particularly important for the production,operation and economic benefits of petrochemical enterprises.Thesis,the combustion status and influencing factors of the heating furnace are studied in order to improve the thermal efficiency of the heating furnace,and predict the thermal efficiency of the heating furnace for a period of time in the future,so that the furnace adjustment decision can be made in advance according to the combustion status of the heating furnace At the same time,aiming at the problem that the prediction effect of traditional neural network is not accurate,a new prediction model is established to improve the prediction accuracy of thermal efficiency.The main research contents are as follows:Firstly,the data related to the thermal efficiency of the heating furnace in a petrochemical enterprise are collected,and the input variables related to the thermal efficiency are analyzed and selected for normalization.The thermal efficiency prediction of this data set is carried out in two ways: support vector regression(SVR)model with strong nonlinear regression ability and DBN-DNN model constructed by deep confidence network(DBN).The simulation results show that the prediction accuracy of the two models is relatively high,but the SVR model is slightly higher than that of DBN-DNN model,because the regression ability of SVR in the output layer is higher than that of BP layer in the output layer of DBN-DNN model.At the same time,in order to obtain higher prediction accuracy,depth confidence network(DBN)is used to extract sufficient features from the data.Considering that in the prediction of heating furnace thermal efficiency,SVR model has stronger generalization ability,and the output BP layer of DBN-DNN model has weaker regression ability to data than SVR,so SVR is selected as the output layer of the model.In view of the problem that particle swarm optimization algorithm has slow convergence speed and is easy to fall into local optimal when optimizing SVR parameters,an improved particle swarm optimization algorithm is proposed to optimize SVR parameters as the DBN-SVR model based on improved particle swarm optimization algorithm.The experimental results show that the new prediction model improves the prediction accuracy obviously.Secondly,according to the factors affecting the thermal efficiency of the heating furnace analysis,put forward the corresponding solutions.The heating furnace combustion model was established and GAMBIT was used for modeling.According to the comparison of simulation experiments,P-1 model,which can better experience the combustion state,was selected as the radiation model.Considering that the index of oxygen content is the most critical factor affecting the combustion and thermal efficiency,the optimal control of the combustion is proved to be less than 3%,and the optimal thermal efficiency can be achieved.Finally,in order to better verify the new prediction model established in this paper,the model is embedded in the monitoring platform of a petrochemical enterprises,and the trial operation shows that the prediction model can well predict the heating furnace thermal efficiency in the future period of time,so as to make an outstanding contribution to the optimization of combustion and improvement of petrochemical production.