为实现对醇类柴油的鉴别和含量的测定, 在实验室制备了甲醇柴油和乙醇柴油。利用主成分分析方法将甲醇柴油和乙醇柴油分成两类, 利用中红外光谱法对光谱数据进行平滑、基线校正、多元散射校正和归一化, 得到预处理光谱图, 使用最小二乘支持向量机(LSSVM)模型预测甲醇柴油的甲醇含量和乙醇柴油的乙醇含量, 其模型误判率低于7.1%。结果表明, 甲醇柴油甲醇含量LSSVM模型预测集相关系数和预测集均方误差分别为0.9791和1.7201; 乙醇柴油乙醇含量LSSVM模型预测集相关系数和预测集均方误差分别为0.9802和2.9563。实验的结果表明最小二乘支持向量机模型能够预测甲醇乙醇柴油中的甲醇和乙醇含量, 且是一种很有前途的方法。

In order to realize the identification and content determination of methanol/ethanol diesel, the methanol diesel and ethanol diesel are prepared in the laboratory and divided into two groups by principal component analysis. Based on the mid-infrared spectra method, the preprocessing spectra are produced by smoothing, baseline correction, multiplicative scatter correction and normalization of the spectral data. The least squares support vector machine (LSSVM) prediction model is used to predict methanol/ethanol content, and the error rate is lower than 7.1%. The simulated results show that the correlation coefficient and the mean square error of the LSSVM prediction for methanol content of methanol diesel are 0.9791 and 1.7201; the correlation coefficient and root mean square error are 0.9802 and 2.9563 for ethanol content of ethanol diesel, respectively. The experimental results show that the proposed model is a promising choice to identification and content determination of methanol/ethanol diesel.