Comprehensive Assessment of Windings Deterioration in Oil-Immersed Transformers Based on Stacked Autoencoder

To achieve a more accurate assessment of the current deterioration level of transformers, a comprehensive deterioration indicator based on a SAE network was developed. Firstly, deterioration indicators were constructed based on long-term influencing factors such as temperature and load ratio. Subsequently, deterioration indicators were separately constructed based on current mechanical vibration signals, internal insulation defects, historical fault information of the transformer, and other factors. The SAE network was then employed to intelligently reduce the dimensionality of the multidimensional information, resulting in a comprehensive deterioration indicator. Case study results demonstrate that the proposed method accurately reflects the current deterioration status of transformer windings.