Abstract: As a critical component in the power system, transformers play a vital role in ensuring the stable operation of the electrical grid. The identification of loose winding conditions in transformers holds significant importance. Addressing issues related to traditional monitoring methods, such as susceptibility to environmental interference and complexity of application, a novel approach is proposed. This approach involves the utilization of two different types of fiber bragg grating（FBG） sensors to collect temperature and strain signals at key points within the transformer windings. After the data is processed through fast decoupling and complete ensemble empirical mode decomposition with adaptive noise（DECE）, essential parameters are extracted and subjected to principal component analysis（PCA）. The reduced-dimensional features are then classified using support vector machine based on black hole optimization（BHOSVM）, enabling the monitoring and localization of radial loosening in transformer windings. The diagnostic method proposed achieves an accuracy rate of 96.8% in identifying the radial loosening condition of transformer windings.