摘要
针对电站锅炉燃烧效率经典建模方法精度较低的问题,提出了一种基于卷积神经网络(CNN)的新型建模方法。通过分析影响锅炉燃烧效率的各项因素,建立了以锅炉负荷、一二次风配风方式、烟气含氧量等参数为输入,锅炉燃烧效率为输出的建模数据集。由于CNN的核心特征是通过卷积运算从样本数据的空间关联中进行特征提取,所以为了加强输入样本数据的空间特征,利用主成分分析法(PCA)对数据进行重构。在此基础上,采用LeNet-5结构的CNN构建锅炉燃烧效率模型。仿真结果表明,经过PCA重构的锅炉燃烧效率CNN模型的检验集相对误差最小,相比经典模型减小了19.58%,该方法为锅炉燃烧效率的优化提供了新的思路。
A novel modeling approach based on a convolutional neural network(CNN)is proposed to solve low precision problem of the classical modeling approaches for combustion efficiency of power plant boilers A data set for modeling is established through analyzing factors affecting boiler combustion,and operation parameters including the boiler load,the primary and secondary air distribution,the oxygen content in flue gas are taken as inputs and the combustion efficiency is an output.Since the core characteristic of CNN is to extract features from the spatial association among sample data by convolution,the input data are reconstructed by principal component analysis(PCA)method to enhance its spatial features.On this basis,the CNN with LeNet-5 structure is adopted to build the boiler combustion efficiency model.Simulation results and comparisons with some classical modeling methods show that the relative error of the proposed model for boiler combustion efficiency on testing set is minimal,and the relative error reduces by 19.58%.This novel modeling approach provides a new idea for the optimization of boiler combustion efficiency.
作者
胡怀中
张建博
刘汉青
李梦迪
杨清宇
HU Huaizhong;ZHANG Jianbo;LIU Hanqing;LI Mengdi;YANG Qingyu(School of Electronic and Information Engineering,Xi’an Jiaotong University,Xi’an 710049,China)
出处
《西安交通大学学报》
EI
CAS
CSCD
北大核心
2019年第10期10-15,共6页
Journal of Xi'an Jiaotong University
基金
国家自然科学基金资助项目(61673315)
关键词
电站锅炉燃烧效率
卷积神经网络
特征提取
主成分分析法
power plant boiler combustion efficiency
convolutional neural networks
feature extraction
principal component analysis
