摘要
为了准确预测热电厂锅炉的NO_x排放质量浓度,以某热电厂300 MW亚临界循环流化床锅炉为研究对象,利用并联型快速学习网进行建模。首先以经典的回归数据集测试并联型快速学习网的有效性,将结果与其他神经网络的结果相比较,证明其具有更好的学习能力和稳定性。再以热电厂现场采集的样本数据作为模型的输入输出数据,将该模型的预测结果与极限学习机、快速学习网、核极限学习机和增量型极限学习机的预测结果进行比较。结果表明:并联型快速学习网具有良好的预测精度和泛化能力,能够更准确有效地预测热电厂锅炉的NO_x排放质量浓度。
To accurately predict the NOx emission from boilers of thermal power plants,a 300 MW subcritical circulating fluidized bed（CFB）boiler was taken as the object of study,for which a model was established using fast learning network with parallel layer perceptron.Firstly,the validity of the proposed network was tested using classical regression data sets,and the results were subsequently compared with other neural networks,so as to examine its learning ability and stability.Then the sample data collected from the thermal power plant were taken as the input and output variables of the model proposed.Finally,the prediction results were compared with those of extreme learning machine,fast learning network,kernel extreme learning machine and incremental extreme learning machine,etc.Results show that the model established based on fast learning network with parallel layer perceptron has a good prediction accuracy and generalization ability,which may accurately and effectively predict the NOxemission from boilers of thermal power plants.
作者
李国强
齐晓宾
陈彬
张露
LI Guoqiang , QI Xiaobin , CHEN Bin, ZHANG Lu(Key Lab of Industrial Computer Control Engineering of Hebei Province, Yanshan University, Qinhuangdao 066004, Hebei Province, Chin)
出处
《动力工程学报》
CAS
CSCD
北大核心
2018年第6期440-446,共7页
Journal of Chinese Society of Power Engineering
基金
国家自然科学基金资助项目(61403331
61573306)
河北省自然科学基金资助项目(F2016203427)
关键词
NOx排放质量浓度
循环流化床锅炉
并联型快速学习网
极限学习机
NOx emission concentration
circulating fluidized bed boiler
fast learning network with parallel layer perceptron
extreme learning machine
