为了快速、准确地测定强化生物除磷(EBPR) 过程中污泥胞内糖原的含量,采用 4种预处理方法分别对污泥近红外光谱进行预处理,并结合联合区间偏最小二乘(siPLS) 进行变量优选,建立光谱吸光度数据与糖原含量的定量分析模型。结果表明,将一阶 S-G(Savitzky-Golay) 平滑处理后的光谱等分为 20个子区间,联合子区间[10 13 16 19]建立的 siPLS模型预测效果最优,预测集的均方根误差(RMSEP) 和相关系数(rp) 分别达到 0.0048、0.9105,且该模型的交互验证和外部验证相对分析误差(RPD) 均大于 3.0。一阶平滑处理后的光谱 siPLS模型预测精度高、建模变量少,可实现糖原含量的快速测定。

In order to develop a procedure to determine the intracellular glycogen content in the sludge from the enhanced biological phosphorus removal (EBPR) process rapidly and accurately, four Near-infrared spectral pre-processing methods were separately used to improve the performance of multivariate calibrations. The quantitative analysis models between the near-infrared (NIR) spectra and the glycogen contents were established based on synergy interval partial least squares(siPLS) method. Results showed that the optimal model was established by four sub-intervals[10 13 16 19]with siPLS method after the NIR spectra were treated with Savitzky-Golay smoothing and first derivative as pre-processing method, and the preprocessed spectra were devided into 20 sub-intervals. The root mean square errors (RMSEP) and the correlation coefficients (rp) of the prediction sets were 0.0048 and 0.9105 respectively. The residual predictive deviation (RPD) values of cross-validation and prediction sets were all above 3.0. The NIR spectral model based on S-G smoothing and siPLS method can not only improve the predictive accuracy and reduce the variable numbers, but also realize the rapid and accurate determination of intracellular glycogen content in the sludge.