Partial Discharge Spectral Characterization in HF, VHF and UHF Bands Using Particle Swarm Optimization.

Area of Science:

• Electrical Engineering
• Signal Processing
• Electromagnetic Compatibility

Background:

• Partial discharge (PD) measurements in radio frequency (RF) bands are crucial for electrical asset monitoring.
• Noise and interference often obscure weak PD signals, particularly in ultra-high frequency (UHF) bands, hindering accurate interpretation.
• Existing dimensionality reduction techniques can lose critical physical signal characteristics, complicating analysis.

Purpose of the Study:

• To develop a selective spectral signal characterization method for improved PD signal analysis.
• To create a preprocessing tool that reduces data dimensionality by focusing on representative frequency bands.
• To enhance the clustering of PD signals and differentiate them from noise and interference.

Main Methods:

• Utilizing selective spectral characterization to condense signal energy into reduced frequency bands (UHF, HF/VHF).
• Applying dimensionality reduction techniques to preserve signal characteristics for easier interpretation.
• Employing particle swarm optimization (PSO) to maximize the minimum Mahalanobis distance between signal clusters, optimizing frequency band selection.

Main Results:

• Demonstrated capability in separating noise from PD signals, even those with low signal-to-noise ratios.
• Successfully differentiated between various types of partial discharges.
• Validated the method's effectiveness across UHF and HF/VHF bands using experimental data.

Conclusions:

• The proposed selective spectral characterization is an effective preprocessing tool for PD monitoring systems.
• The method enhances the separation of PD signals from noise and interference.
• This technique improves the ability to distinguish between different PD types, crucial for accurate fault diagnosis.