浙江电力

2022, v.41;No.311(03) 93-99

[打印本页] [关闭]
本期目录(Current Issue) | 过刊浏览(Archive) | 高级检索(Advanced Search)

基于神经网络的CFB锅炉灰渣含碳量预测
Prediction of Carbon Content in CFB Ash Based on Neural Networks

陈斌,王树宇,刘林涛,朱伟,董瑀非
CHEN Bin,WANG Shuyu,LIU Lintao,ZHU Wei,DONG Yufei

摘要(Abstract):

建立神经网络子母模型用于实时预测CFB(循环流化床)锅炉灰渣平均含碳量,其中子模型通过煤质参数得到煤质燃尽特性指数,母模型通过锅炉运行参数与煤质燃尽特性指数预测灰渣含碳量的实时值,训练后子模型和母模型的MAE(平均绝对误差)分别为0.343 2和0.974 7。将某热电厂现场CFB锅炉性能试验数据代入神经网络模型进行计算,得到灰渣平均含碳量预测值与真实值的MAE为0.84%,锅炉效率预测值与真实值的MAE为0.15%,证明模型具有较好的泛化性。最后,建立一个直接将煤质参数和锅炉运行参数作为输入参数的神经网络模型,训练后模型的MAE为1.038 0,灰渣含碳量和锅炉效率的预测值与真实值的MAE分别为1.29%和0.23%,误差大于使用子模型的预测误差,验证了子模型的必要性。
The parent/sub-model of a neural network is established to monitor the average carbon content of CFB(circulating fluidized bed) ash in real time. The sub-model obtains the coal burnout characteristic index through the coal quality parameters,and the parent model predicts the real-time value of ash carbon content through the boiler operation parameters and coal burnout characteristic index. After training,the MAE(mean absolute error) of submodel and parent model are 0.343 2 and 0.974 7 respectively. By substituting the on-site boiler performance test data of a thermal power plant into the neural network model,the MAE between the predicted value and the real value of average carbon content in ash slag is 0.84%,and the MAE between the predicted value and the real value of boiler efficiency is 0.15%,which demonstrate the good generalization. Finally,a neural network model using coal quality parameters and boiler operation parameters as input parameters is established. The post-training MAE is1.0380,and the MAEs of predicted values and real values of boiler efficiency and carbon contents of the ash are1.29% and 0.23% respectively,and the errors are larger than those of the sub-model,which verifies the necessity of the sub-model.

关键词(KeyWords): 循环流化床锅炉;煤质燃尽特性指数;人工神经网络;灰渣含碳量
CFB;coal burnout characteristic index;artificial neural network;carbon content in ash

Abstract:

Keywords:

基金项目(Foundation):

作者(Author): 陈斌,王树宇,刘林涛,朱伟,董瑀非
CHEN Bin,WANG Shuyu,LIU Lintao,ZHU Wei,DONG Yufei

DOI: 10.19585/j.zjdl.202203012

参考文献(References):

扩展功能
本文信息
服务与反馈
本文关键词相关文章
本文作者相关文章
中国知网
分享