基于LightGBM-CatBoost的循环流化床锅炉NOx排放浓度预测研究
武洁 , 周会成 , 梅文 , 刘小恺 , 云达娜 , 王峰 , 唐忠锋 , 张志勇
现代化工 ›› 2026, Vol. 46 ›› Issue (6) : 244 -251.
基于LightGBM-CatBoost的循环流化床锅炉NOx排放浓度预测研究
Prediction of NOx emission concentration in circulating fluidized bed boiler based on LightGBM-CatBoost.
为实现燃煤机组变负荷下循环流化床(Circulating Fluidized Bed,CFB)锅炉的NOx排放浓度预测,采用数据驱动-集成学习的4种算法构建了6个预测模型。以某300 MW燃煤机组的CFB锅炉为例,采用6个模型算法研究了不同负荷、给煤量和一次风量对CFB锅炉NOx排放浓度的影响,并对其NOx排放浓度进行了预测。研究结果表明,与其他5个模型相比,LightGBM-CatBoost(L-C)模型的平均绝对误差和均方根误差分别为0.12和0.16,拟合优度为0.98,预测精度和稳定性最佳。随CFB锅炉的负荷变化,NOx排放浓度预测值与实测值最大相对误差小于13.9%,预测值相对误差小于3%。给煤量变化时,XGBoost-CatBoost和LightGBM-NGBoost 2模型预测的NOx排放浓度值与实测值最大相对误差为9%,小于L-C模型预测的相对误差13.9%。在NOx浓度高于120 mg/m3时,L-C模型预测的NOx排放浓度值与实测值最大偏差小于5%。L-C模型能够保证宽负荷条件下对CFB锅炉NOx排放浓度值精准预测。本研究可以为选择性非催化还原技术的低氮氧化物燃烧及控制提供新的思路,对CFB锅炉的灵活性改造提供了重要的方法和数据支持。
NOx emission concentration of circulating fluidized bed (CFB) boilers were predicted by six prediction models consisted four data-driven ensemble learning algorithms.The effects of load,coal feed and primary airflow on NOx emission concentration were investigated and a 300 MW CFB boiler was employed as an example.Compared with the other five models,the LightGBM-CatBoost (L-C) model displayed the best prediction accuracy and stability with an MAE and RMSE of 0.12 and 0.16,which R-squared was 0.98.The maximum relative error between the predicted and measured NOx emission concentration was less than 13.9% with the changes of load,and most of relative error of the predicted values was less than 3%.The maximum relative error was 9% between the NOx emission concentration values predicted by the XGBoost-CatBoost and LightGBM-NGBoost models and the measured values when the coal feed varies,which was smaller than the relative error of 13.9% predicted by the L-C model.The maximum deviation of NOx emission concentration values predicted by the L-C model was less than 5% when the measured values of NOx emission concentrations were higher than 120 mg/m3.The L-C model ensures the accurate prediction of NOx emission concentration for CFB boiler under the condition of wide load.It provides an advanced theory for selective non-catalytic reduction technology in low NOx combustion control and supports theoretical foundations and data for the implementation of flexibility retrofits in CFB boilers.
数据驱动 / 循环流化床 / NOx / 集成学习 / 烟气脱硝
data-driven / circulating fluidized bed / NOx / ensemble learning / flue gas denitrification
| [1] |
唐晶, 欧阳昊东, 孙文涛, |
| [2] |
李月娥, 吴彦丽, 郑立星, |
| [3] |
蒋海威, 高明明, 张开萍, |
| [4] |
张文祥, 晏海能, 孙志军, |
| [5] |
田继林, 朱建国, 朱书骏, |
| [6] |
吕清刚, 朱书骏, 朱建国, |
| [7] |
杨希刚, 陈国庆, 黄林滨, |
| [8] |
欧阳子区, 王宏帅, 吕清刚, |
| [9] |
|
| [10] |
|
| [11] |
任少君, 朱保宇, 翁琪航, |
| [12] |
|
| [13] |
|
| [14] |
|
| [15] |
|
| [16] |
|
| [17] |
|
| [18] |
苏连成, 朱娇娇, 郭高鑫, |
| [19] |
张子洵, 魏业文, 张轲钦, |
| [20] |
陈林, 陶润哲, 谢雨池, |
| [21] |
|
| [22] |
|
| [23] |
|
| [24] |
|
| [25] |
蔡晋, 单露, 王志宁, |
| [26] |
|
中央引导地方科技发展资金项目(2023ZY0006)
内蒙古电力集团(有限)责任公司内蒙古电力科学研究院分公司自筹科技项目(2024-ZC-2-04)
/
| 〈 |
|
〉 |