基于间隔偏最小二乘法短程硝化反硝化中无机盐氮的近红外光谱

采用序批式活性污泥反应器(SBR)研究短程硝化反硝化系统中稳定条件下无机盐氮的变化规律,利用小波去噪法对无机盐氮的近红外光谱进行预处理,并用间隔偏最小二乘法(iPLS)建立无机盐氮含量的校正模型(iPLS模型).结果表明:小波去噪法对原始光谱中的部分噪声进行滤除,从而提高了模型的精度.采用最小二乘法对预处理后的光谱进行建模,优选出最佳波长区间并将光谱划分为20个子区间,优选出的氨氮特征波数为8243~8663cm-1,亚硝酸盐氮特征波数为4000~4420cm-1.所建模型对氨氮、亚硝酸盐氮校正时的相关系数(rc)分别达到0.9582、0.9544,校正均方根误差(RMSECV)分别为0.0321、0.0406;预测时的相关系数(rp)分别为0.9184、0.8816,预测均方根误差(RMSEP)分别为0.0790、0.0451.采用实际污水为原水时校正模型对氨氮、亚硝酸盐氮预测的相关系数(rp￠)分别为0.9190、0.8739,预测均方根误差(RMSEP′)分别为0.0578、0.0229.模型对氨氮和亚硝酸盐氮的预测总体效果较好.用小波去噪和最小二乘法建立模型不仅能有效减少建模所需变量数、缩短运算时间,而且模型预测精度较高,为无机盐氮的快速测定提供了一种可行的分析技术.

Near infrared spectroscopy analysis of inorganic nitrogen in shortcut nitrification-denitrification based on interval partial least square

The change law of the inorganic nitrogen in the stable shortcut nitrification and denitrification systems was researched with sequencing batch activated sludge reactor (SBR). Wavelet denoising method was used to preprocess the near infrared spectra of inorganic nitrogen and interval partial least square (iPLS) method was used to establish the mathematical prediction models of inorganic nitrogen content. The results indicated that wavelet denoising method could partially remove the noise of original spectra and improve the accuracy of the model. The interval partial least squares (iPLS) was used to model the preprocessed spectra, to decide the optimum wavelength range and divide the spectrum into 20sub-interval. The optimum wave number range of the ammonia was 8243~8663cm-1while the nitrite nitrogen was 4000~4420cm-1.The correction coefficients of the models for the ammonia nitrogen and nitrite nitrogen were 0.9582 and 0.9544, respectively, and root mean square errors of cross validation (RMSECV) were 0.0321 and 0.0406, respectively, and the prediction coefficients were 0.9184 and 0.8816 with the root mean square errors of prediction (RMSEP) of 0.0790 and 0.0451. When real wastewater was used as raw water in the process, the prediction coefficients of the models for the ammonia nitrogen and nitrite nitrogen were 0.9190 and 0.8739, respectively, and root mean square errors of cross validation (RMSECV) were 0.0578 and 0.0229. The prediction effects of the model on ammonia nitrogen and nitrite nitrogen were better. The method of wavelet denoising and iPLS can effectively not only reduce the variable numbers for establishing models and short operation time, but increase the accuracy of the model, thus providing a feasible analysistechnique for the rapid determination of inorganic nitrogen.