Research on Nonuniform Ageing State Assessment of Oil‐Immersed Power Transformers Based on PCA–GWO–RF Algorithm

ABSTRACT

The condition assessment of oil–paper insulation is critical for oil–immersed power transformers, which are fundamental to the reliability of modern power systems. The primary limitation of traditional transformer ageing assessment lies in its macroscopic, whole‐unit approach, which fails to account for the localised effects of nonuniform ageing (NUA) induced by temperature gradients. To address this, the influence of NUA must be explicitly incorporated into the evaluation process. An equivalent scaled‐down model representing the main insulation of an oil–immersed power transformer was constructed in this paper. Frequency‐domain dielectric response (FDS) measurements were then performed on the scaled model, as well as on oil–paper insulation (OPI) samples with different ageing states, including uniform and nonuniform ageing. The results reveal the FDS properties of nonuniformly aged OPI. Considering the geometric structure of an oil–immersed power transformer, an equivalent capacitance model was developed to characterise the NUA of the main insulation, which exists between windings with different voltage levels. Furthermore, a quantitative analysis method for the NUA state of the main insulation based on the principal component analysis (PCA)–grey wolf optimiser (GWO)–random forest (RF) algorithm was proposed, and the assessment results were experimentally validated. The maximum relative error of the assessment results was less than 6%. The findings of this study provide theoretical support for developing operation and maintenance strategies for the power system.