Abstract: This paper proposes a novel approach for assessing the insulation condition of transformers by combining the non-dominated sorting genetic algorithm Ⅱ（NSGA-Ⅱ） with an extended Cole-Cole model. The method is applied to extract feature parameters for the evaluation of insulation states in laboratory samples and transformers. Initially, frequency domain spectroscopy messages are collected from insulation samples in different states. Subsequently, uncoupling analysis is performed on the extended Cole-Cole model to distinguish various dielectric response behaviors. The NSGA-Ⅱ algorithm is then utilized to solve the extended Cole-Cole model and separate the feature parameters, which are further refined using the Pearson correlation coefficient method. Finally, based on the obtained feature parameters, insulation state assessments are conducted on laboratory samples and on-site transformers using neural network tools. The results demonstrate that this intelligent algorithm-based approach, coupled with frequency domain spectroscopy, has high accuracy and applicability in evaluating the insulation condition of transformer oil-paper insulation.