FAN Yuchen, ZHOU Yongqing, WEI Chang, et al. Data-driven Modeling for Over-temperature Prediction of Platen Superheater in Coal-fired Boiler[J]. 2025, 45(19): 7634-7643.
FAN Yuchen, ZHOU Yongqing, WEI Chang, et al. Data-driven Modeling for Over-temperature Prediction of Platen Superheater in Coal-fired Boiler[J]. 2025, 45(19): 7634-7643. DOI: 10.13334/j.0258-8013.pcsee.241152.
燃煤锅炉普遍采用空气分级燃烧技术,此举虽可大幅降低NOx生成,但造成炉内火焰中心上移,导致屏式过热器(屏过)管壁超温严重。此外,调峰运行使锅炉负荷经常性不规则变化,进一步恶化了屏过传热,使爆管泄漏事故频发。为指导锅炉安全可靠运行,提出一种基于遗传算法优化超参数的深度神经网络模型(deep neural network model with its hyperparameters optimized by genetic algorithm,GA-DNN),通过构建炉内风煤侧及汽水侧运行参数与屏过30片管屏出口温度之间的映射关联,对屏过超温进行分析和预测。该模型可实现对不同负荷工况下屏过温度分布的准确预测,在此基础上能够以97.5%以上的准确率识别出当前及未来5 min屏过超温(> 550℃)的运行工况,同时可在89.2%的准确率下预测出未来5 min屏过超温最严重的管屏所在区域。
Abstract
Air-staging combustion technology is widely used in coal-fired boilers to reduce NOx emissions. However
it causes high-temperature flame to move upwards in the furnace and aggravates the tube over-temperature of the platen superheater (PLSH) located in the upper furnace. Moreover
the rapid changes of unit load under load-cycling mode further aggravate the tube over-temperature of boiler PLSH. In order to guide the safe and reliable operation of boilers
a deep neural network model with hyperparameters optimized by genetic algorithm (GA-DNN) is proposed to predict the over-temperature of boiler PLSH. By establishing the mapping correlation between the air
coal
and steam parameters and 30 tube panel temperatures at PLSH-outlet
the model can accurately predict the temperature distribution of the PLSH under different load conditions. On this basis
it can identify the over-temperature operational conditions (> 550℃) of PLSH for the present and for the next 5 min with over 97.5% precision. Additionally
it can predict the region with the worst PLSH in the next 5 min