灵活性深度调峰下锅炉NO_x排放的神经网络方法预报  

Neural Network Prediction of NO_x Emission Characteristics of Boiler under Flexibility and Deep Peak Shaving

作　　者：景雪晖 张涛 周曙明 许雷 JING Xuehui;ZHANG Tao;ZHOU Shuming;XU Lei(State Grid Xinjiang Electric Power Research Institute,Urumqi 830001,China;Changji Thermoelectricity Huadian Xinjiang Power Co.,Ltd,Changji 831100,China)

机构地区：[1]国网新疆电力有限公司电力科学研究院,新疆乌鲁木齐830001 [2]华电新疆发电有限公司昌吉热电厂,新疆昌吉831100

出　　处：《上海电力学院学报》2019年第3期215-220,共6页Journal of Shanghai University of Electric Power

基　　金：国网新疆电力有限公司科技项目(大型火力发电机组深度调峰技术研究,5230DK17000R)

摘　　要：利用MATLAB软件中自带的神经网络算法模块对经典文献所载数据和方法进行了校核。在该方法的基础上,将某330MW机组在深度调峰期间低负荷下的运行数据作为已知数据,就地实测的选择性催化还原技术(SCR)入口NOx排放值作为输出值,采用经典的Levenberg-Marquardt训练算法,建立了神经网络训练模型。训练结果表明,输出值和目标值的拟合R值接近0.98,MATLAB软件自带的神经网络算法可以预报SCR入口NOx的排放值,实现了在深度调峰低负荷运行期间达到降低试验工作量、减少试验成本的目的。The data and methods of the classical literature are checked by the neural network algorithm module in Matlab.On the basis of this method,the operation data of a 330 MW unit under the low load during the depth peak adjustment is taken as the known data,the NO x emission values are measured at selective catalytic reduction(SCR)entrance,and the classical Levenberg-Marquard is used.The neural network training model is established.The results show that the R value of the output value and the target value fitting is close to 0.98.The neural network algorithm in Matlab can predict the NO x emission value at the selective catalytic reduction(SCR)entrance,and realize the purpose of reducing the test workload and the test cost during the low load operation of the depth peak regulation.

关 键 词：深度调峰 锅炉 神经网络 

分 类 号：TM621.2[电气工程—电力系统及自动化]