NO_x Emission Of Coal-fired Power Plant Boiler Model And Optimization Based On Artificial Neural Network

The regulations of reducing coal-fired boiler NOx emissions become more and more strict.To reduce coal-fired boiler NOx emissions under the condition of guaranteeing the boiler efficiency. This article adopts the method of combustion optimization adjustment, mainly done the following work:Based on the scene of a ultra supercritical 660 MW coal-fired power station boiler thermal state experiment data samples, the BP neural network and RBF neural network and the SVM(support vector machine) neural network and Elman neural network method regression algorithms were used respectively to establish models which can control the NOx emissions characteristics of a coal-fired boiler by using intelligent MATLAB toolbox. The momentum method is adopted to improve the problems existing in the BP neural network. Optimizing the RBF neural network prediction model of output layer weights, RBF function deviation between center and standard value, Elman network model for the selection of the corresponding parameters; The SVM prediction model for the kernel function and the corresponding parameters selection in c and g; Modeling results show that the four kinds of modeling method has better accuracy and generalization ability. The research results show that the four kinds of modeling method have better accuracy and generalization ability.Choosing the most commonly used three kinds of model including BP network model, RBF(radial basis) network model and the SVM(support vector machine) network model from the above four models.Compared the simulation and prediction results of BP model with the other two kinds of model. It is concluded that the SVM(support vector machine) network model and RBF network model are superior to the BP network model in terms of computing speed, fit, and generalization ability under the circumstances of less boiler thermal state data samples. The SVM(support vector machine) network model has the best performance.There are many defects in the application of neural network algorithm, such as easily falling into local minimum problem during the process of global search, its structure design also has certain defects, such as network convergence speed is slow. Using genetic algorithm to optimize weights and threshold of BP network model, can realize the optimization of the model, the prediction accuracy and generalization ability of optimized model have improved.Using genetic algorithm to optimize parameters including coal mill, the secondary air damper opening, burning wind throttle opening and export oxygen chamber of a stove or furnace on the basis of the BP network model. This paper adopts the multi-objective function in a certain target as the optimization goal, other objective functions are limited within a certain range of the method, the multi-objective optimization problem is transformed into single objective optimization problem. Solving a certain load of the boiler under low NOx emissions by limiting the certain range of the boiler combustion efficiency. Optimization results show that the combustion optimization method, you can find the boiler efficiency of give attention to two or morethings and NOx emissions of optimal operation parameters.