TL;DR
This paper introduces a novel PRPD-based algorithm for filtering and reducing noise in partial discharge (PD) signals from gas-insulated switchgear (GIS). By analyzing PD characteristics within the PRPD (phase-resolved partial discharge) representation, the algorithm effectively suppresses interference without distorting the essential PD signal, leading to improved detection sensitivity and accuracy.
What Problem Does This Paper Address?
Detecting partial discharges (PD) in gas-insulated switchgear (GIS) is crucial for maintaining the reliability of electrical equipment. However, PD signals are often weak and easily obscured by significant electrical noise and interference present in the operational environment. This noise can originate from various sources, including power supplies, electromagnetic interference, and poor connections. The presence of this noise makes it challenging to accurately identify and classify different types of PD, potentially leading to missed fault detection and equipment failure. Existing noise reduction methods often struggle to effectively separate PD signals from noise without losing critical signal characteristics or introducing new distortions.
Methods
The core of this paper is a "secondary noise reduction" algorithm that leverages the Phase-Resolved Partial Discharge (PRPD) representation. Instead of directly analyzing the raw signal, the algorithm first converts the PD signal into a pulse format and then extracts key features from the PRPD, such as the effective signal and pulse width.
The algorithm employs two filtering strategies:
- Periodic Filtering: This targets periodic interference by identifying and masking signals that fall within a specific range of effective signal intervals. The algorithm calculates an average spacing over eight cycles to establish a periodic parameter.
- Non-periodic Filtering: This addresses non-periodic interference by specifying and masking signals based on their pulse width. The pulse width is determined relative to the sampling rate.
These two filtering methods are combined to create a masking effect that aims to remove interference while preserving the PD signal's characteristics. The algorithm then regenerates the PRPD mapping with the reduced noise.
Additionally, the paper describes the design of a dual-channel quadrature down-converter and a dual-channel data acquisition system to improve signal processing and anti-interference capabilities.
Datasets
Not discussed in the paper. The experiments were conducted using a self-designed GIS PD fault simulation system and on-site testing.
Key Insights and Ideas
- PRPD as a Feature Extraction Tool: The paper highlights the power of PRPD not just for visualization but as a rich source of features (pulse width, effective signal) for targeted noise reduction.
- "Secondary Noise Reduction": The approach focuses on refining the signal after initial acquisition, assuming that some level of noise is unavoidable. This is a practical strategy for real-world scenarios.
- Dual Filtering for Diverse Noise: Combining periodic and non-periodic filtering strategies allows the algorithm to tackle a broader spectrum of interference types commonly found in GIS environments.
- Preserving Signal Integrity: A key goal is to remove noise without distorting the PD signal's inherent characteristics, which is crucial for accurate PD type identification.
Key Results
- Improved Signal-to-Noise Ratio (SNR): The proposed PRPD filtering algorithm significantly improved the SNR. For example, conventional noise reduction yielded an SNR of 7.96 dB, while the PRPD algorithm achieved 13.98 dB.
- Enhanced Sensitivity: The system's sensitivity, measured by the minimum transient electric field strength peak (Eimin), was improved from 1.8 V/m (conventional noise reduction) to 1.138 V/m (with PRPD secondary noise reduction). A lower Eimin indicates higher sensitivity.
- Accurate PD Type Identification: The noise reduction algorithm did not interfere with the identification of five typical PD types (tip discharge, particle discharge, air gap discharge, suspension discharge, and surface discharge), demonstrating that essential signal characteristics were preserved.
- Effective On-Site Performance: Field tests confirmed that the algorithm effectively filtered out noise in a noisy environment, resulting in a clean signal map.
Limitations and Future Work
- Limitations: The paper acknowledges that complete noise elimination while perfectly retaining signal characteristics is not always achievable due to various environmental, equipment, and material limitations. The current method is described as "secondary noise reduction" and does not analyze the noise source itself.
- Future Work: The authors suggest that future work will involve conducting an adaptive study of PRPD threshold parameters. This implies that the current thresholds might be fixed and could benefit from dynamic adjustment based on real-time signal conditions.
Summary updated