基于变量选择的燃煤机组SCR脱硝系统SSA-KELM建模  

SSA-KELM Modeling of SCR Denitration System for Coal-fired Units Based on Variable Selection

作　　者：姜浩 许子明 赵文杰[1] JIANG Hao;XU Ziming;ZHAO Wenjie(Department of Automation,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学自动化系,河北保定071003

出　　处：《电力科学与工程》2023年第11期71-78,共8页Electric Power Science and Engineering

摘　　要：针对燃煤机组选择性催化还原(Selective catalytic reduction,SCR)脱硝系统出口NOx浓度存在测量滞后以及吹扫时数据失真等问题,提出基于特征提取和麻雀搜索算法(Sparrow search algorithm,SSA)优化核极限学习机(Kernel based extreme learning machine,KELM)超参数的燃煤机组SCR脱硝系统建模方法。利用最大信息系数确定各输入变量的延迟时间,表征变量间的相关性。在此基础上采用相关性的特征选择算法将加入迟延参数的重构输入变量进行变量选择;通过SSA优化算法确定KELM初始输入层权值及偏差的超参数。所建立的SSA-KELM预测模型的均方误差和相关系数分别为1.1212 mg/m3、0.9616,与粒子群算法、灰狼算法寻优后的预测模型所得结果相比预测精度较高,模型能够为脱硝系统出口NOx的现场优化控制提供技术支持。Aiming at the problems of NOx concentration at the exit of selective catalytic reduction(SCR)denitration system in coal-fired units,such as the measurement lag and the data distortion during purging,a hyper parameter modeling method for SCR denitrification system of coal-fired units is proposed based on feature extraction and sparrow search algorithm(SSA)to optimize the kernel extreme learning machine(KELM).The delay time of each input variable is determined by the maximum information coefficient,and the correlation between variables is characterized.On this basis,a feature selection algorithm of correlation is used to select the reconstructed input variables with delay parameters,and the hyper parameters of KELM’s initial input layer weights and deviation are determined by SSA optimization algorithm.The mean squared error and correlation coefficient of the SSA-KELM prediction model are 1.1211.1211.1211.1211.1212 mg/m2 mg/m2 mg/m2 mg/m2 mg/m 3 and 0.9616 respectively.Compared with the prediction model optimized by particle swarm optimization and gray wolf algorithm,the prediction accuracy is higher,and the model can provide technical support for the on-site optimal control of NOx at the outlet of the denitrification system.

关 键 词：SCR脱硝系统 最大信息系数 麻雀搜索算法 相关性的特征选择算法 核极限学习机 

分 类 号：TK39[动力工程及工程热物理—热能工程]