基于RBF神经网络和BP神经网络的燃煤锅炉NO_x排放预测  

NO_x emission prediction for coal-fired boilers based on RBF Neural network and BP Neural network

作　　者：余廷芳[1] 刘冉[1] 

机构地区：[1]南昌大学机电工程学院,江西南昌330031

出　　处：《热力发电》2016年第8期94-98,113,共6页Thermal Power Generation

基　　金：国家自然科学基金资助项目(61262048)

摘　　要：基于某超超临界660 MW机组燃煤锅炉现场热态实验数据,利用MATLAB智能工具箱,分别采用径向基(RBF)神经网络和BP神经网络对该锅炉NOx排放特性进行建模,采用交替梯度算法对RBF神经网络预测模型进行输出层权值及RBF函数的中心与标准偏差值优化,对BP神经网络采用动量法进行改进。2种模型的仿真和预测结果对比分析表明:参数优化后的RBF神经网络预测模型预测结果的最大误差为3.0%,平均误差为1.75%;改进后的BP神经网络预测模型预测结果最大误差为6.6%,平均误差为4.5%;2种建模方法均具有较好的准确性和泛化能力,其中RBF神经网络模型的计算速度快,拟合和泛化能力更强。Combing with the hot state test data of an ultra supercritical 660 MW coal-fired boiler,the radial basis function（RBF）neural network and BP neural network regression algorithms were used to establish models for NOxemissions from this boiler,by using the intelligent MATLAB toolbox.Moreover,optimization was performed for the RBF neural network prediction model of output layer weights and the center of RBF function and standard deviation,and the momentum method was adopted to improve the problems existing in the BP neural network.The results show that,the maximum error of the RBF prediction model after parameter optimization simulation was 3.0% and the average error was 1.75%.The maximum error of the improved BP network prediction model was 6.6%and the average error was 4.5%.Both the above two modeling methods have well accuracy and generalization ability.The RBF neural network model is obviously superior to the BP neural network model in terms of computing speed,fitting ability and generalization ability.

关 键 词：燃煤锅炉 NOX排放 RBF神经网络 BP神经网络 预测模型 

分 类 号：TK223.23[动力工程及工程热物理—动力机械及工程]