Accurate diagnosis of oil-paper insulation material is an essential technology to ensure the safe operation of equipment insulation. Raman spectroscopy has been widely used in material composition analysis and state diagnosis. In this paper, the oil-paper insulation samples obtained from the accelerated thermal ageing experiment were divided into four ageing stages according to the degree of polymerization of the paper, and researched based on the Raman spectroscopy platform. By analyzing the energy information contained in the Raman spectra of different ageing samples, the feature vector was extracted by Wavelet Packet Energy Entropy, and the discriminant function was constructed by Fisher discriminant method. What’s more, the ageing model of oil-paper insulation based on Raman spectroscopy was established, and oil samples of running transformers were collected to verify the generalization ability of diagnostic model. The results show that two discriminant functions can distinguish different ageing stages, and the accuracy rate of ageing samples reaches 84.2%. Raman spectroscopy combined with wavelet packet entropy and Fisher discriminant analysis can effectively diagnose the ageing state of oil-paper insulation.