In order to respectively assess the aging degree and moisture content of transformer oil-paper insulation

this article prepares oil impregnated insulation paper samples with different moisture content and aging degree through accelerated thermal aging test and moisture absorption experiment

then tests the dielectric loss factor spectrum and the argument of geometric impedance spectrum of different samples

and analyzes the influence of moisture and aging to frequency domain dielectric spectroscopy.Finally

adopted multi-output support vector regression algorithm（M-SVR）to approximate the nonlinear relationship between frequency dielectric spectrum curve moisture content and aging degree

then accordingly predict the aging degree and moisture content of the samples.Research results show that for different aged oil-paper samples with moisture from 0.5 to 5.6%

the prediction accuracy of moisture content is relatively high

and the prediction accuracy for aging degree is not acceptable

but both of the accuracy is higher than RBF neural network.According to the clustering analysis result to all samples

the study finds that for the samples belonging to class 1 and 2 which is more correlative to aging

the aging prediction accuracy by M-SVR increase obviously

while the moisture prediction accuracy changed a little.