中文摘要：

为了进一步提高多组分气体分析的准确度，对采用AOTF—NIR光谱仪采集甲烷、乙烷和丙烷多组分混合气体的近红外光谱数据建立了新的分析模型。首先对光谱数据采用偏最小二乘法（以下简称PLS）进行特征提取，随后将提取得到的潜变量作为支持向量回归机（以下简称SVR）的输入建立多组分混合气体的定量分析模型。结果显示，PLS特征提取耦合SVR对近红外光谱的定量分析取得了很好的分析效果。

英文摘要：

In this paper, the quantitative analysis of multi-component gas mixture of methane, ethane, and propane was done on AOTF-NIR spectrometer. In order to improve the prediction accuracy, a new analysis model was built. Firstly, partial least squares （PLS） method was used for feature extraction of spectral data, and extracted latent variables were input into the support vector regression （SVR） to establish a quantitative analysis model of multi-component gas mixture. The research show that PLS feature extraction coupled with SVR made a good analysis result for the quantitative analysis of near infrared spectroscopy.