To achieve quantitative analysis of Raman spectra of the characteristic gases in transformer oil

the molec-ular structure optimization and vibration model of seven dissolved gases in transformer oil was set up. Based on thesecond-order Moller-Plesset perturbation theory（MP2）on 6-31 G（d）basis set

the bond length

bond angle and spec-tral peak ascription information of the seven gases’ s molecules after optimization were obtained. On the Raman d

etec-tion test platform

the Raman detection of the seven characteristic gas samples with same concentrations and the CH4 samples with different concentrations was completed

and the spectral information of H

2

CO

CO

2

CH

4

C

2

H

2

C

2

H

4

and C

2

H

6

was obtained. Combining with the molecular Raman vibration theory

a method for quantitative analy-sis based on multiple Raman peak data at different frequency shifts of polyatomic molecules was put forward.Through multi-parameter regression

the equation of the relationship between the CH

4

spectrum’s peak areas at thefrequency shifts of 2 878 cm

-1

and 2 982 cm

-1

and the gas concentration was achieved. The reliability of the moleculartheory model and the multi-parameter regression approach is verified

which may provide a new idea for quantitativeanalysis of Raman spectrum in transformer fault diagnosis.