Assessment of UHF Frequency Range for Failure Classification in Power Transformers.

Area of Science:

• Electrical Engineering
• Materials Science
• Condition Monitoring

Background:

• Ultrahigh-frequency (UHF) sensing is crucial for detecting partial discharges (PDs) in power transformer insulation.
• Uncertainty exists regarding the optimal UHF frequency range for accurate fault diagnosis.
• Bandwidth reduction may impact the performance of frequency-based failure classification systems.

Purpose of the Study:

• To assess the impact of spectrum band limitation on power transformer failure classification.
• To evaluate the effectiveness of signal processing techniques based on UHF signal frequency content.
• To investigate the feasibility of using reduced frequency bands for cost-effective data acquisition.

Main Methods:

• Three power transformer operational conditions (healthy, electric arc, bushing PDs) were analyzed.
• Three self-organized maps (Chromatic Technique, PCA, SACT) were used for failure classification.
• UHF signal frequency content was analyzed across three bands: full spectrum, 100 MHz-1 GHz, and 400 MHz-900 MHz.

Main Results:

• Reduced frequency ranges did not significantly impair the classification of transformer operational conditions.
• The 400 MHz-900 MHz band proved effective for failure classification.
• Principal Component Analysis (PCA) demonstrated the highest promise among the evaluated techniques.

Conclusions:

• Lower UHF frequency ranges, such as 400 MHz-900 MHz, are viable for power transformer fault classification.
• Reduced bandwidth monitoring systems can be developed, leading to cost savings.
• PCA is a robust technique for UHF-based transformer condition monitoring, even with limited frequency data.