Furfural content in transformer oil is one of the important indexes of insulation aging in power transformers. Infrared and Raman spectroscopy have the advantages of fast detection speed, fast analysis speed and non-destructive detection in effectively identifying substances. This paper compares furfural detection methods in transformer oil based on infrared and Raman spectroscopy. Based on the Fourier transform infrared spectrometer and laser confocal Raman spectroscopy instrument configuration of furfural transformer sample testing for the laboratory and its spectral data, using wavelet transform, the multinomial least square method and locally weighted regression fitting the collected infrared and Raman spectra is preprocessed, careful consideration of noise, resolution and effective information loss, etc., The polynomial least square method has the best preprocessing effect. The molecular model of furfural was established based on Gaussian software, and the attribution of the infrared and Raman absorption peaks of furfural was studied by density functional simulation calculation. The Raman and infrared detection characteristic peaks of furfural in transformer oil were determined to be 1 703 and 1 704 cm-1, respectively, combined with the experimental spectra. The repeatability experiments of infrared and Raman detection were carried out, and the relative standard deviations of the two detection methods were 7.21% and 8.67%, respectively. The relationship between the infrared and Raman 3D in situ spectrographs was analysed by drawing furfural infrared and Raman characteristic peak areas and transformer oil with different furfural concentrations. The infrared and Raman quantitative analysis models for furfural detection in transformer oil were established based on the least square method, and the goodness of fit was 0.998 and 0.885, respectively. Compared to the two kinds of detection methods of quantitative analysis of the model prediction results, the results show that Raman spectroscopy detection lower limit is lower than the infrared spectrometry, pretreatment by the multinomial least square method of spectral data of transformer oil furfural quantitative analysis of infrared and Raman model can well predict furfural content in transformer oil, to provide technical reference for related research.