摘要
以某600 MW旋流燃烧器锅炉燃烧优化调整时集散控制系统(DCS)数据选取特性参数作为输入变量,训练神经网络预测该燃煤锅炉氮氧化物排放量。使用单隐含层BP神经网络、双隐含层BP神经网络、遗传算法优化神经网络、广义回归(GRNN)神经网络作为预测网络模型,分析预测误差。对比各模型均方根误差(RMSE)与平均绝对百分比误差(MAPE),GRNN神经网络预测精度最高,可用于对该机组进行氮氧化物排放量预测。
Using the distributed control system(DCS) data of a 600 MW swirl burner boiler during combustion optimization and adjustment to select characteristic parameters as input variables, training a neural network to predict the nitrogen oxide emissions of the coal-fired boiler. The single hidden layer BP neural network, double hidden layer BP neural network, genetic algorithm optimization neural network, generalized regression(GRNN) neural network were selected to analyze prediction errors. Comparing the root mean square error(RMSE) and mean absolute percentage error(MAPE) of each model, the GRNN neural network has the highest prediction accuracy, and can be used to predict nitrogen oxide emissions for this unit.
出处
《节能》
2022年第8期22-26,共5页
Energy Conservation
关键词
混煤
神经网络
氮氧化物排放量预测
blended coal
neural network
prediction of nitrogen oxide emission