TL;DR
This paper introduces a novel zero-sequence differential current protection scheme for converter transformers that uses waveform correlation analysis to accurately distinguish between internal and external faults, even in the presence of challenging conditions like CT saturation and recovery inrush currents. This improved scheme enhances the reliability of power systems by preventing false trips and ensuring proper fault detection.
What Problem Does This Paper Address?
Converter transformers are critical components in ultra-high voltage direct current (UHVDC) transmission systems, connecting AC and DC grids. Zero-sequence differential current protection is a vital protective measure for these transformers, particularly for detecting grounding faults. However, existing protection schemes face significant challenges:
- Recovery Inrush Current: When an external fault is removed, a transient "recovery inrush current" can flow. This current, especially when combined with the inherent characteristics of converter transformers, can lead to saturation of the neutral current transformer (CT).
- CT Saturation: CT saturation distorts the measured current signals, leading to a false differential current. This distortion can cause the traditional zero-sequence differential protection to mis-operate, incorrectly tripping the system during non-fault conditions or failing to operate during actual faults.
- External Fault Persistence: Even without recovery inrush, external faults can cause CT saturation, leading to similar mis-operation risks.
These issues compromise the reliability of the protection system, potentially leading to power outages and equipment damage. The paper aims to address these vulnerabilities by developing a more robust protection scheme.
Methods
The core innovation of this paper is a zero-sequence differential current protection scheme based on waveform correlation analysis. Here's how it works:
- Zero-Sequence Current Measurement: The scheme monitors the zero-sequence currents at two key locations: the terminal of the converter transformer (denoted as $3I_{self0}$) and the neutral point of the transformer ($3I_{no}$).
- Waveform Correlation: Instead of solely relying on the magnitude difference of these currents (as in traditional schemes), this new method analyzes the similarity between the waveforms of $3I_{self0}$ and $3I_{no}$ over a specific time window.
- Correlation Coefficient Calculation: A correlation coefficient ($r_{ab}$) is calculated between the two zero-sequence current signals. This coefficient ranges from -1 to 1, where:
- A value close to 1 indicates high similarity (strong positive correlation), suggesting an external fault where currents are largely in phase.
- A value close to -1 indicates a strong dissimilarity (strong negative correlation), suggesting an internal fault where currents are largely out of phase.
- Values between -1 and 1 indicate varying degrees of correlation.
- Improved Identification Criterion: The protection scheme uses a dual criterion:
- The differential current ($I_{op} = |3I_{self0} - 3I_{no}|$) must exceed a predefined operating threshold ($I_{op.0}$).
- The calculated correlation coefficient ($r_{ab}$) must be below a set threshold ($r_{ab.set}$). This threshold is specifically chosen to differentiate between internal and external faults, even under challenging conditions.
This approach leverages the fact that external faults (and their removal, under normal CT operation) tend to produce similar zero-sequence current waveforms at both measurement points, while internal faults create significantly different waveforms. The correlation analysis effectively amplifies these differences, making it easier to distinguish between fault types.
Datasets
The paper does not use traditional "datasets" in the machine learning sense. Instead, it relies on:
- Simulations: Extensive simulations were performed using the PSCAD/EMTDC software. This is a standard tool for electromagnetic transient analysis in power systems.
- ±800 kV UHVDC Transmission Model: A detailed simulation model of a ±800 kV UHVDC transmission system was established, including a converter transformer (T1) on the rectification side. This model incorporates realistic parameters for the transformer and its associated components, including a J-A transformer model for the neutral point CT to accurately represent saturation characteristics.
The simulations covered various fault scenarios, including:
- External faults (with and without CT saturation)
- Recovery inrush currents (with CT saturation)
- Internal faults (with and without CT saturation)
Key Insights and Ideas
- Waveform Correlation as a Robust Discriminator: The most significant insight is that waveform correlation is a powerful tool for distinguishing between internal and external faults, especially when CT saturation distorts current magnitudes. The phase relationship between the zero-sequence currents is more resilient to saturation effects than their amplitude.
- Recovery Inrush is a Major Threat: The paper clearly demonstrates that recovery inrush currents, particularly when combined with CT saturation, pose a substantial risk of mis-operation for traditional differential protection schemes.
- CT Saturation Mitigation: The proposed correlation-based criterion exhibits a notable ability to resist CT saturation. By focusing on the waveform's shape and phase, it can still correctly identify fault types even when the measured current magnitudes are distorted.
- Unified Approach for Multiple Fault Types: The new criterion effectively handles various fault scenarios, including external faults, internal faults, and the problematic recovery inrush conditions, all within a single, unified framework.
Key Results
The simulation results, verified across multiple fault scenarios, demonstrate the effectiveness of the proposed waveform correlation-based protection scheme:
- Accurate Internal/External Fault Discrimination: The scheme reliably distinguishes between internal and external faults, achieving a correlation coefficient close to -1 for internal faults and close to 1 for external faults (under normal CT operation).
- Robustness Against CT Saturation: In cases of external faults with CT saturation, the traditional scheme mis-operated (indicated by 'x' in Table 2), while the new criterion maintained a high correlation coefficient (e.g., 0.96), preventing mis-operation.
- Effective Handling of Recovery Inrush: During recovery inrush with CT saturation, the traditional scheme mis-operated (0.51 p.u. differential current), but the new criterion maintained a correlation coefficient of 0.57, which was then successfully discriminated against by the set threshold of 0.4, preventing mis-operation.
- Reliable Internal Fault Operation: For internal faults, the correlation coefficient consistently dropped to -1, ensuring proper protection action.
- Set Threshold Optimization: The paper proposes a critical correlation coefficient threshold ($r_{ab.set}$) of 0.4, which was found to be effective in distinguishing between fault types across various simulated conditions.
Limitations and Future Work
- Simulation-Based Verification: While the simulations are extensive and use a detailed model, the proposed scheme's performance in real-world power systems with all their complexities still needs to be validated through field testing.
- Parameter Tuning: The optimal threshold ($r_{ab.set}$) was determined based on the ±800 kV UHVDC model. Further research might be needed to generalize this threshold for different voltage levels or transformer configurations.
- Computational Load: While not explicitly detailed as a limitation, real-time implementation of correlation analysis might introduce a slightly higher computational load compared to simpler magnitude-based methods, which could be a consideration for very fast protection requirements.
Future work could involve:
- Field testing of the proposed scheme.
- Investigating the impact of different CT models and saturation characteristics.
- Exploring adaptive threshold setting for the correlation coefficient based on system conditions.
- Extending the application of this waveform correlation technique to other types of differential protection.
Summary updated