广义回归神经网络对脱硫效率的预测  

Prediction of Desulfurization Efficiency Based on Generalized Regression Neural Network

作　　者：李军红[1] 刘锁清[1] 董森[1] 彭伟娟 刘少虹 LI Jun-hong;LIU Suo-qing;DONG Sen;PENG Wei-juan;LIU Shao-hong(Shanxi University,Taiyuan 030013 China)

机构地区：[1]山西大学,山西太原030013

出　　处：《自动化技术与应用》2018年第10期1-3,37,共4页Techniques of Automation and Applications

摘　　要：脱硫效率的预测对脱硫系统的运行与控制有着重要的指导意义。以国内某660MW机组为例,考虑影响石灰石/石膏湿法烟气脱硫效率的各主要因素,并使用广义回归神经网络(GRNN)建立了脱硫效率预测模型。该模型采用电厂实际运行数据为训练样本,然后另选10组数据用来仿真预测和验证。预测结果表明建立的烟气脱硫效率预测模型的精度要高于传统BP模型,精度达到了99.6%,对实际脱硫系统的安全运行有一定的指导意义。The prediction of desulfurization efficiency has important guiding significance for the operation and control of desulfurization system.Taking a 660MW supercritical unit as an example,the main factors that affect the efficiency of limestone-gypsum wet flue gas desulfurization are taken into account,the prediction model which is based on generalized regression neural network(GRNN)is built.The model uses the actual running data of power plant as training samples,and then selects 10 sets of data to predict and verify by simulation.The simulation results show that the established prediction model of flue gas desulfurization efficiency has higher accuracy than the traditional BP model,and the accuracy is 99.6%.So it has certain guiding significance for the operation of the actual desulfurization system.

关 键 词：脱硫效率 广义回归神经网络 预测模型 训练样本 仿真预测 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]