水泥分解炉SNCR脱硝系统的深度强化学习多目标优化控制研究  

Research on Deep Reinforcement Learning Multi-objective Optimization Control of SNCR Denitration System of Cement Calciner

作　　者：刘定平[1] 吴泽豪 LIU Dingping;WU Zehao(Guangdong Province Engineering Research Center of High Efficiency and Low Pollution(South China University of Technology),Guangzhou 510640,Guangdong Province,China)

机构地区：[1]广东省能源高效低污染转化工程技术研究中心(华南理工大学),广东省广州市510640

出　　处：《中国电机工程学报》2024年第12期4815-4825,I0017,共12页Proceedings of the CSEE

摘　　要：选择性非催化还原(selective non-catalytic reduction,SNCR)脱硝过程的工艺参数优化可以有效减少水泥分解炉NO_(x)排放和脱硝运行成本。以某水泥分解炉为研究对象,建立基于LightGBM的NO_(x)浓度预测模型,以脱硝成本和NO_(x)浓度最小化为优化目标,采用深度确定性策略梯度(deep deterministic policy gradient,DDPG)算法对水泥分解炉掺烧污泥协同SNCR脱硝过程的相关工艺参数进行优化控制建模。结果表明,NO_(x)浓度预测模型均方根误差(root mean squared error,RMSE)为6.8,平均绝对百分比误差(mean absolute percentage error,MAPE)为3.48%;采用DDPG算法可以对相关工艺参数进行优化,喷氨量和污泥掺烧量分别为427.87 L/h和9.78 t/h时,NO_(x)排放浓度为225.99 mg/(Nm^(3)),脱硝运行成本为1 747.8元/h。该优化结果与其他优化算法结果和常规工况对比,NO_(x)排放浓度和脱硝运行成本均呈现不同程度下降;对模型进行仿真及效果验证可知,所建立模型能输出合理的喷氨量和污泥掺烧量组合,减少SNCR出口NO_(x)浓度波动,有效降低NO_(x)排放浓度和脱硝成本,可实现对SNCR脱硝系统的多目标优化控制。该结果可为基于智能算法的水泥分解炉SNCR脱硝的多目标优化控制设计提供一定参考。The process parameter optimization of selective non-catalytic reduction(SNCR)denitration process can effectively reduce NO_(x) emission and denitration cost of cement calciner.Taking a cement calciner as the research object,a NO_(x) concentration prediction model based on LightGBM is established.Taking the minimization of denitration cost and NO_(x) concentration as the optimization objective,the deep deterministic policy gradient(DDPG)algorithm is used to optimize and control the relevant process parameters of SNCR denitration process of cement calciner mixed with sludge.The results show that the RMSE of NO_(x) concentration prediction model is 6.8 and MAPE is 3.48%.The DDPG algorithm can effectively optimize the relevant process parameters:when the ammonia injection amount and sludge combustion amount are 427.87 L/h and 9.78 t/h,respectively,the NO_(x) emission concentration is 225.99 mg/(Nm^(3))and the denitration operation cost is 1747.8 yuan/h.Compared with the results of other optimization algorithms and conventional working conditions,the NO_(x) emission concentration and denitration cost of DDPG optimal control model show different degrees of decline.The simulation and effect verification of the model show that the model can output a reasonable combination of ammonia injection and sludge combustion,reduce the fluctuation of NO_(x) concentration at the outlet of SNCR,effectively reduce NO_(x) emission concentration and denitration cost,and realize the multi-objective optimal control of SNCR denitration system.This study can provide a reference for the multi-objective optimal control design of SNCR denitration of cement calciner based on intelligent algorithm.

关 键 词：喷氨 污泥掺烧 选择性非催化还原优化控制 LightGBM 强化学习 深度确定性策略梯度 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置]