The aging degree of an oil-paper insulation system is closely related to the stable operation of power transformer equipment. There is low accuracy in aging diagnosis when it relies on a single relevant feature quantity obtained by frequency-domain spectroscopy (FDS) and does not consider the conflict and randomness among the spectral indicators. Thus a new diagnostic method for determining the aging degree of oil-paper insulation is proposed. This method integrates spectral characteristics and clustering cloud-evidence reasoning. First, the multi-feature indicators for aging diagnosis are extracted through aging samples’ spectral characterization. An aging feature database is then constructed through nonlinear fitting. Secondly, a basic probability assignment method for the clustering cloud model is proposed based on K-means clustering to accommodate the nonlinear change indicators. Finally, to address the conflict and correlation differences of multi-feature quantities, the CRITIC-G1 comprehensive assignment method is used to calculate the evidence correction factor for each indicator. Subsequently, the basic probability is reassigned, and evidence fusion inference is employed to determine the actual aging degree of the system. The results demonstrate that the proposed method can accurately diagnose the aging state of composite oil-paper insulation samples and offers theoretical guidance for developing power maintenance strategies.