Early Fault Detection of High-Voltage Bushings Based on Curve Recognition Methods

The aim of this study is the early detection of developing insulation defects in high-voltage oil-filled bushings based on the analysis of the temporal behavior of diagnostic indicators rather than only their comparison with threshold values. To achieve this aim, insulation indicators are interpreted as time curves, and data processing is performed using curve recognition methods that identify charac-teristic patterns of parameter variation. The most important results demonstrate stable differences between different technical conditions. Serviceable bushings are characterized by the absence of a statistically significant relationship with service duration while maintaining consistent internal dependencies and a high similarity of time curves on adjacent phases caused by common operating conditions. For defective bushings, statistically significant relationships between individual indica-tors and service duration are identified, changes in the internal correlation structure are observed, and similarity with the indicators of neighbouring serviceable bushings is absent. Based on the recognition of the shapes and mutual consistency of time curves, a decision rule is formulated: a defect is diagnosed when correlation with time and internal correlations are present, while cross-phase correlation is simultaneously absent. The significance of the obtained results lies in the fact that the application of curve recognition methods enables the detection of developing defects before the indicators exceed threshold values, reduces the probability of erroneous rejection, and increases the reliability of bushing condition assessment under various network operating modes. The pro-posed approach can be used in technical condition monitoring systems to improve the efficiency of operational maintenance of high-voltage equipment, leading to reduced failure risks and more justified maintenance planning.