摘要
锅炉燃烧过程属于持续性工艺流程,当前运行工况参数会受到前N个周期的工况叠加影响。本文收集锅炉负荷、省煤器出口氧量、各二次风挡板开度、燃尽风挡板开度、各磨煤机给煤量、炉膛与风箱差压、一次风总风压、锅炉运行中排出的煤灰和煤渣的含碳量等参数,形成时间序列样本集,构建LSTM神经网络模型,用于预测燃煤锅炉热效率。该方法能够挖掘并记忆锅炉连续运行过程中参数自身变化与热效率影响的客观规律,克服锅炉持续性燃烧调整的工况叠加带来的预测误差,提高学习效率,提升预测精度。
The boiler combustion is a continuous technological process,and the current operating conditions parameters will be affected by the superposition of the first N cycles.This paper collected boiler parameters such as boiler load,economizer outlet oxygen,secondary air baffle opening,burnout baffle opening,coal supply to each coal mill,furnace and bellows differential pressure,primary wind total pressure,carbon content of coal ash and cinder discharged during boiler operation.Then a time series sample set was formed and an LSTM neural network model was constructed to predict the thermal efficiency of the coal-fired boiler.The method could excavate and memorize the objective law of the influence of parameters itself and the thermal efficiency during the continuous operation of the boiler.It also overcome the prediction error brought by the superposition of the continuous combustion adjustment of the boiler,improved the learning efficiency and the prediction accuracy.
作者
李佳鹤
徐慧
张静
周献军
LI Jia-he;XU Hui;ZHANG Jing;ZHOU Xian-jun(Hangzhou Deep Blue Digital Technology Co.,Ltd.,Hangzhou 310053,China;Zhejiang Dahua Technology Co.,Ltd.,Hangzhou 310053,China)
出处
《智能物联技术》
2019年第3期33-36,共4页
Technology of Io T& AI
关键词
LSTM神经网络
时间序列
锅炉
热效率预测
LSTM neural network
time series
boiler
thermal efficiency prediction