烟气中SO_2紫外光谱的加性噪声去除及评价方法

Denoising and Assessing Method of Additive Noise in the Ultraviolet Spectrum of SO_2 in Flue Gas

用紫外差分吸收光谱法对烟气中主要污染物SO2测量中的噪声去除及信噪比评价的问题进行了研究。在获取被测气体吸收特征时,采用基于多分辨率的原始光谱预处理方法,在不同尺度下判断信号的能量幅度滤除加性噪声,并根据烟气光谱信号在时间序列上不会出现突变的特性,在含有吸收特征的尺度上提高有用信号强度。再使用吸收截面构造理想吸收信号改善光谱的信噪比评价。新的去噪方法分别在实验室和现场进行了验证。在实验室中,使用上述方法对SO2气体进行多次测量,平均偏差均小于1.5%,重复性不超过1%。在现场测量中,以一个气体浓度变化范围较大的山东电厂为例,在18组比对数据中,最大偏差为2.31%。实验结果说明,该方法可有效地提高加性噪声污染严重的光谱的信噪比。

The problem of denoising and assessing method of the spectrum of SO2 in flue gas was studied based on DOAS. The denoising procedure of the additive noise in the spectrum was divided into two parts：reducing the additive noise and enhancing the useful signal. When obtaining the absorption feature of measured gas,a multi-resolution preprocessing method of original spectrum was adopted for denoising by DWT （discrete wavelet transform）. The signal energy operators in different scales were used to choose the denoising threshold and separate the useful signal from the noise. On the other hand,because there was no sudden change in the spectra of flue gas in time series,the useful signal component was enhanced according to the signal time dependence. And the standard absorption cross section was used to build the ideal absorption spectrum with the measured gas temperature and pressure. This ideal spectrum was used as the desired signal instead of the original spectrum in the assessing method to modify the SNR （signal-noise ratio）. There were two different environments to do the proof test-in the lab and at the scene. In the lab,SO2 was measured several times with the system using this method mentioned above. The average deviation was less than 1.5%,while the repeatability was less than 1%. And the short range experiment data were better than the large range. In the scene of a power plant whose concentration of flue gas had a large variation range,the maximum deviation of this method was 2.31% in the 18 groups of contrast data. The experimental results show that the denoising effect of the scene spectrum was better than that of the lab spectrum. This means that this method can improve the SNR of the spectrum effectively,which is seriously polluted by additive noise.