摘要
针对现有加热炉炉膛内三维温度场重建方法存在的重建误差较大、重建消耗时间较长的问题,提出基于正则化SVD算法的660MW机组煤粉加热炉炉膛三维温度场重建方法。根据人眼视觉二维图像特征点提取原理,提取温度场立体图像特征点;利用小波变换方法计算子线段端点,获取特征点匹配结果;通过声学测温方法以及射线成像理论,重建声波传播速度分布形式,凭借正则化SVD算法构建声学测量系统模型,对声波飞行值进行修正,结合特征点匹配结果和对称轴,得到实现660MW机组煤粉加热炉炉膛三维温度场重建。实验结果表明,所提方法的最低AER、MER、RMSE分别为4.11、0.98、1.21,重建时间始终保持在0.6s以内,重建误差较小、重建消耗时间较短,抗噪声能力强,温度场重建效果好。
A regularized SVD algorithm based method for reconstructing the three-dimensional temperature field of a 660 MW coal-fired heating furnace furnace is proposed to address the problems of large reconstruction errors and long reconstruction time in existing methods for recon-structing the three-dimensional temperature field in the furnace.Based on the principle of extracting feature points from two-dimensional ima-ges of human vision,extract feature points from three-dimensional images of temperature fields;Using wavelet transform method to calculate the endpoints of sub line segments and obtain feature point matching results;The acoustic temperature measurement method and the ray ima-ging theory are used toreconstruct the distribution form of the acoustic wave propagation velocity.The acoustic measurement system modelis built by using the regularization SVD algorithm to modify the acoustic wave flight value.Combining the feature point matching results and the symmetry axis,realized the three-dimensional temperature field reconstruction of the coal powder heating furnace furnace of the 660 MW unit.The experimental results show that the minimum AER,MER,and RMSE of the proposed method are 4.11,0.98,and 1.21,respectively,and the reconstruction time remains within O.6 seconds.The reconstruction error is small,the reconstruction consumption time is short,the noise resistance is strong,and the temperature field reconstruction effect is good.
作者
陈鹏
邢军
孙黎君
CHEN Peng;XING Jun;SUN Lijun(Huayuan Power Plant of State Grid Energy Hami Coal Power Co.Ltd.,Hami 839000,China)
出处
《工业加热》
CAS
2024年第5期58-63,共6页
Industrial Heating
基金
国网能源哈密煤电有限公司(GSKJ-22-05)。
关键词
燃烧温度
正则化SVD算法
特征点提取
三维温度场重建
小波变换
combustion temperature
regularization SVD algorithm
feature point extraction
reconstruction of three-dimensional temperature field
wavelet
