采用紫外可见光谱(UV-Vis)与极限学习机算法检测水体化学需氧量(chemical oxygen demand, COD)含量研究。 采集135份水样进行紫外可见波段全光谱扫描, 结合变量标准化(standard normal variate, SNV), 多元散射校正(MSC)和一阶微分(1st D)对原始数据进行预处理, 然后采用竞争性自适应重加权算法(competitive adaptive reweighted sampling, CARS)、 随机青蛙(Random frog)算法和遗传算法进行特征波长选择。 基于全光谱建立了偏最小二乘回归(partial least squares, PLS)和基于特征波长建立了极限学习机算法(extreme learning machine, ELM)模型。 结果表明: 使用CARS提取的9个特征波长建立的ELM模型的预测效果最优, 决定系数R2为0.82, 预测均方根误差RMSEP为 14.48 mg·L-1, RPD值为2.34。 说明使用CARS变量选择算法获取UV-Vis光谱特征波长, 应用极限学习机建模, 可以准确、 快速的检测养殖水体中COD含量, 为实现养殖水体COD的动态快速检测以及水体其他微量物质含量参数检测打下基础。

Ultraviolet/visible (UV/Vis) spectroscopy technology was used to measure water COD. A total of 135 water samples were collected from Zhejiang province. Raw spectra with 3 different pretreatment methods (Multiplicative Scatter Correction (MSC), Standard Normal Variate (SNV) and 1 st Derivatives were compared to determine the optimal pretreatment method for analysis. Spectral variable selection is an important strategy in spectrum modeling analysis, because it tends to parsimonious data representation and can lead to multivariate models with better performance. In order to simply calibration models, the preprocessed spectra were then used to select sensitive wavelengths by competitive adaptive reweighted sampling (CARS), Random frog and Successive Genetic Algorithm (GA) methods. Different numbers of sensitive wavelengths were selected by different variable selection methods with SNV preprocessing method. Partial least squares (PLS) was used to build models with the full spectra, and Extreme Learning Machine (ELM) was applied to build models with the selected wavelength variables. The overall results showed that ELM model performed better than PLS model, and the ELM model with the selected wavelengths based on CARS obtained the best results with the determination coefficient (R2), RMSEP and RPD were 0.82, 14.48 and 2.34 for prediction set. The results indicated that it was feasible to use UV/Vis with characteristic wavelengths which were obtained by CARS variable selection method, combined with ELM calibration could apply for the rapid and accurate determination of COD in aquaculture water. Moreover, this study laid the foundation for further implementation of online analysis of aquaculture water and rapid determination of other water quality parameters.