基于Q-learning和随机森林的SCR脱硝系统温度优化研究  

Research on Temperature Optimization of SCR Denitrifi cation System Based on Q-learning and Random Forest

作　　者：李玮[1] Li Wei(Ironworks of Shanghai Meishan Iron and Steel Co.,Ltd.,Jiangsu,210039)

机构地区：[1]上海梅山钢铁股份有限公司炼铁厂,江苏210039

出　　处：《当代化工研究》2024年第19期155-157,共3页Modern Chemical Research

基　　金：梅山钢铁设备部科研资助项目“皮带机安全消防及状态提升”(项目编号:W24AAP1)。

摘　　要：钢铁行业的烧结工艺会产生大量污染源,如重金属颗粒物、NO_(x)等。为了控制污染物浓度,需要采用脱硫、除尘、脱硝等工艺对烟气进行净化处理。低温SCR脱硝工艺在钢铁行业中应用广泛,但由于温度控制依赖人工经验,存在时滞、多变量等问题,导致煤气过量消耗和运行成本增加。本文针对这一问题,提出了一种结合大数据与人工智能的优化控制方法,通过引入Q-learning算法,结合随机森林模型预测的煤气用量,对低温SCR脱硝系统的入口温度进行优化控制。实验结果显示,优化后的SCR系统入口平均温度为194.2℃,较人工控制的实际平均温度209.8℃明显降低;每月煤气耗量由202.3×10^(4)Nm^(3)降至179.6×10^(4)Nm^(3),节能约9%;氨水耗量由51.7×10^(4)L降至49.2×10^(4)L,节能约5%。研究表明,本文所提出的控制方法能够显著提高脱硝效率,减少煤气消耗,降低生产成本。The sintering process in the steel industry generates a significant amount of pollutants,such as heavy metal particulates and NO_(x).To control the concentration of these pollutants,desulfurization,dust removal,and denitrification processes are employed to purify the flue gas.The low-temperature selective catalytic reduction(SCR)denitrification process is widely used in the steel industry;however,due to the reliance on manual experience for temperature control,issues such as time delays and multivariable complexities arise,leading to excessive gas consump-tion and increased operational costs.To address this problem,this paper proposes an optimized control method that combines big data and artifi-cial intelligence.By introducing the Q-learning algorithm and incorporating the random forest model to predict gas consumption,the inlet tem-perature of the low-temperature SCR denitrification system is optimized.Experimental results show that the optimized SCR system's inlet average temperature is 194.2℃,significantly lower than the manually controlled average temperature of 209.8℃.Monthly gas consumption decreased from 202.3×10^(4)Nm^(3)to 179.6×10^(4)Nm^(3),achieving approximately 9%energy savings.Additionally,ammonia consumption decreased from 51.7×10^(4)L to 49.2×10^(4)L,saving about 5%.The study demonstrates that the proposed control method significantly improves denitrification efficiency,reduces gas consumption,and lowers production costs.

关 键 词：脱硝 SCR系统 随机森林 强化学习 节能减排 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]