SCR脱硝系统PCA-HPSO-SVR大数据建模研究  

Research on the Big Data Modeling of SCR Denitration System Based on PCA-HPSO-SVR

作　　者：高学伟[1,2] 付忠广[1] 刘柄含 张连升 

机构地区：[1]华北电力大学电站设备状态监测与控制教育部重点实验室,北京102206 [2]沈阳工程学院仿真中心,辽宁沈阳110136 [3]国电康平发电有限公司运行部,辽宁沈阳110500

出　　处：《自动化仪表》2017年第7期15-19,共5页Process Automation Instrumentation

基　　金：辽宁省教育厅创新团队基金资助项目(LT2015018)

摘　　要：燃煤火电厂中的选择性催化还原(SCR)脱硝系统无法实现喷氨量的精确控制,SCR脱硝系统模型的可靠性直接关系到脱硝系统控制和运行的可靠性。通过挖掘现场海量的实际运行数据,对采用大数据建模理论建立的脱硝系统模型、SCR脱硝反应系统及现阶段存在的问题进行了研究。通过理论和实际运行情况分析,确定模型输入、输出需采集的变量。采用滑动窗口法提取稳态工况数据,利用主成分支持向量机回归(PCA-SVR)方法进行建模。对标准粒子群算法进行改进,采用混合粒子群算法(HPSO)对模型参数进行优化。结果表明,利用HPSO优化的脱硝系统PCA-SVR模型具有较高的预测精度和泛化能力。所建模型为下一步脱硝系统的优化运行以及喷氨量的精确控制奠定了基础。In coal-fired power plant,the spray ammonia amount of the selective catalytic reduction（ SCR） denitration system cannot be precisely controlled,and the reliability of the model built for SCR denitration system is directly related to the reliability of control and operation of the denitration system. Through the excavation of the massive actual operation data in the field,the denitration system model which is established by big data modeling theory is studied. The SCR denitration reaction system and the problems existing at the present stage are studied,and through analyzing theoretical and actual operation situation,the variables needed to be collected for input and output of the model are determined. The data of steady-state condition are extracted by sliding window method,and the principal component analysis support vector machine for regression（ PCA-SVR） method is used for modeling. The standard particle swarm optimization algorithm（ PSO） is improved,and the model parameters are optimized by hybrid particle swarm algorithm（ HPSO）. The results show that the PCA-SVR model optimized by HPSO offers higher prediction accuracy and better generalization ability. The model established laid the foundation for the optimal operation of the denitration system and the precise control of the spray ammonia amount.

关 键 词：选择性催化还原 大数据 主成分分析 混合粒子群算法 支持向量机回归 滑动窗口法 

分 类 号：TH-3[机械工程] TP39[自动化与计算机技术—计算机应用技术]