以油砂中钠元素为研究对象, 首次应用近红外光谱, 结合Lasso(least absolute shrinkage and selection operator)建模方法, 建立了油砂金属钠含量的近红外光谱定量校正模型, 并与传统的PLS建模方法进行比较。 结果表明, 两种方法建立的油砂金属钠含量校正模型都具有很高的精度, 预测性能方面略有差异。 在实验验证集与预测集中, PLS与Lasso算法的相关系数分别是: Rv=0.878 8, Rp=0.857 9和Rv=0.887 4, Rp=0.860 0。 实验验证了使用近红外光谱快速测定油砂金属钠含量的有效性, 并分析了PLS与Lasso算法的适用范围。

For the sake of the quick analysis of sodium in oil sands, near infrared spectroscopic technology was applied combing with Least Absolute Shrinkage and Selection Operator (Lasso) modeling algorithm in order to establish quantitative calibration model. The comparison with the traditional PLS modeling method was conducted for comparative analysis. The results showed that the calibration models of the sodium content established by both methods had almost the same accuracy, but the prediction performance was slightly different. The verification experiment illustrated that the model evaluation indexes of PLS and Lasso algorithms were Rp=0.998 1, RMSEP=0.010 8 and Rp=0.998 6, RMSEP=0.009 5 respectively. The effectiveness of near-infrared spectroscopic analysis to determine the sodium content in oil sands was verified. The modeling precision and applicable areas of the PLS and Lasso algorithms were compared and analyzed.