基于PSO-BP算法的磨煤机入口一次风量软测量  

Soft Sensing of Primary Air Flow at the Inlet of Coal Mill Based on PSO-BP Algorithm

作　　者：匡世才 KUANG Shi-cai(China Datang Corporation Science and Technology Research Institute Co.,Ltd.,Northwest Branch)

机构地区：[1]中国大唐集团科学技术研究总院有限公司西北电力试验研究院,陕西西安710021

出　　处：《电站系统工程》2023年第6期48-51,共4页Power System Engineering

摘　　要：针对火电机组磨煤机入口一次风量测量误差大、风量自动投入率低的现状,提出一种基于粒子群优化BP神经网络算法的一次风量软测量技术。引入灰色关联性分析法筛选出与待测目标强相关的17维特征参数作为模型输入;通过最近邻法去除样本冗余,构建出10800个高质量样本,采用有限次迭代优选的方式确定最佳隐层神经单元数为4。在某330 MW锅炉上,模型针对2700个测试样本的预测均方根相对误差达到2.30%,满足工业应用需求,为提升锅炉风量自动投入率,加强炉内燃烧调节水平提供了积极有效的探索。Aiming at the large measurement error of primary air flow rate at the inlet of coal mill,which leads to low automatic input rate,proposes a soft measurement technology based on particle swarm optimization BP neural network algorithm.17 dimension characteristic parameters that are strongly related to the target to be measured are selected by introducing the grey relational analysis method as the model input.The nearest neighbor method was used to remove the sample redundancy,and 10800 high-quality samples were constructed.The optimal number of hidden layer neural units was determined to be 4 by finite iteration optimization.On a 330MW boiler,the root mean square relative error of the prediction of the model for 2700 test samples reached 2.30%,meeting the needs of industrial applications,and providing a positive and effective exploration for improving the automatic input rate of boiler air volume and strengthening the combustion regulation level of the boiler.

关 键 词：磨煤机 一次风量 粒子群优化 BP神经网络 

分 类 号：TK313[动力工程及工程热物理—热能工程]