Algorithm for processing transient signals for detection and analysis of partial discharges and insulation condition diagnostics
Abstract
RELEVANCE of the study lies in the need to improve the reliability of medium-voltage distribution networks and the implementation of advanced insulation condition diagnostics based on the analysis of transient process signals (TPS). THE
Purpose
To develop an algorithm for automatic detection and filtering of partial discharge (PD) signals in transient process oscillograms for further diagnostics of insulation condition.
Methods
The proposed algorithm combines median filtering, threshold detection based on the median absolute deviation (MAD), correlation matching with a template library, and local verification of pulse prominence. An exclusion map is integrated into the algorithm to ensure correct removal of PD pulses while preserving the transient process waveform. SCIENTIFIC NOVELTY consists in integrating filtering methods, adaptive threshold detection, and template matching within a unified algorithm for processing TPS, which enables separation of PD pulses from the main signal.
Results
Experimental testing on real oscillograms demonstrated the high sensitivity of the algorithm to short-duration PD pulses, its robustness to noise, and the correct identification of isolated outliers. Diagnostic features were extracted from the processed data, including the number and average amplitude of PD pulses. PRACTICAL
Significance
Oscillograms recorded by traveling-wave fault location (TWFL) systems are formed during scheduled and emergency switching operations in the network, which are accompanied by switching overvoltages. These overvoltages can initiate PD in weakened high-voltage insulation, appearing as short-duration pulses. The proposed algorithm enables extraction of these pulses and thereby allows the already available TWFL measurements to be used for insulation condition diagnostics without installing additional equipment.
Conclusion
The developed algorithm enables efficient detection of partial discharges and the formation of quantitative indicators for indirect assessment of insulation condition. Future work includes expanding the experimental database and implementing intelligent methods for predicting insulation degradation, which will enhance the reliability of medium-voltage distribution networks.