Modeling Analysis Of Boiler Combustion Process Based On Neural Network

The rapid development of modern industry has brought sharp energy consumption,while the boiler is the critical equipment which is related to the industrial productionand the life. Therefore, improving boiler combustion efficiency and reducing emissionsis a significant thing for energy conservation and environmental protection. As theboiler combustion is a complex physical and chemical process, it has greater delay,multivariable coupling and other complex non-linear characteristics, therefore,establishing the exact mechanism model is difficult to achieve, it has become thedifficulty of controlling the boiler combustion.By analyzing the common used modeling methods, BP neural network algorithm isselected for modeling and analyzing the boiler combustion process in this thesis. Thismethod regards the combustion system as a black box, rather than make a concreteanalysis of the internal mechanism of the combustion process. In this context, establisha reasonable model structure. As the BP algorithm has some shortcomings such as slowconvergence rate and easily to fall into the local extremum, we choose the geneticalgorithm to optimize the network and make full use of its global searching ability tooptimize the weights and thresholds of the BP neural network and establish a reasonableGA-BP network model. This algorithm uses the historical operating data to training andtesting the network. From the experimental results we found that using of geneticalgorithm to optimize BP neural network can improve the model accuracy and theconvergence rate, it makes the model doing a better reflect the characteristics of theboiler combustion.In addition, this thesis also analyzes the basic principles of the time series analysisand its major steps to establish a mathematical model about the boiler water temperatureand then test it with the actual historical operating data. By comparing the GA-BP network model and time series models, it can be concluded that the GA-BP networkalgorithm is modeled easily and has a fast convergence characteristic in themulti-variable system; it is more suitable for the modeling of boiler combustion process.In this thesis, the GA-BP network model and the model test were completed byMATLAB. For the time series model, the basic operations of modeling were completedby Eviews.