基于支持向量机的天然气组成多维光谱定量分析方法

针对天然气中甲烷、乙烷、丙烷、正丁烷、异丁烷等气体的特征吸收谱线严重重叠的问题,将支持向量机(SVM)用于天然气组成多维光谱分析中,提出了一种新的基于多维光谱的天然气组成定量分析方法。利用SVM核函数变换,将天然气的多维光谱数据进行高维空间变换后求得SVM回归函数,得出天然气组成。实验结果表明,该方法的检验组成最大偏差为2.28%,与传统的方法相比,具有分析速度快、可实时在线分析、不限制光谱数据维数等优点,具有重要的理论和应用价值。

Method of Natural Gas Composition Quantitative Analysis Based on Support Vector Machine and Multi -Dimensional Spectra

In order to overcome difficulties caused by natural gas with components like CH_4,C_2H_6,C_3H_8,n-C_4H_(10) and i-C_4H_(10) whose characteristic absorption spectra seriously overlapped,a novel quantitative analysis method based on support vector machine and multidimensional spectra was proposed.Through transformation of kernel function,multi-dimensional spectra were mapped into high dimension space,so that component concentrations of natural gas were carried out in high dimension space based on support vector machine regression function.Experimental results of natural gas showed that the component concentration maximal deviations were 2.28%.The method performance was better than traditional method.It could speed up quantitative analysis,make analyzing natural gas online possible and use multi-dimensional spectra data.It was of important theoretical and application value.