toplogo
Sign In

Partial Discharge Ignition in H2 Bubbles within Dielectric Oils: A Numerical Analysis for High-Voltage Solid State Transformer Applications


Core Concepts
The presence of hydrogen gas bubbles, especially those in contact with metallic protrusions, significantly increases the risk of partial discharge ignition in high-voltage transformers, even under typical operating conditions.
Abstract

Bibliographic Information:

Kourtzanidis, K., Dimitrakellis, P., & Rakopoulos, D. (Year). Numerical analysis of partial discharge ignition in H2 bubbles floating in dielectric oils, for High-Voltage Solid State Transformer applications. [Journal Name].

Research Objective:

This study investigates the factors influencing partial discharge (PD) inception in high-voltage solid-state transformers (SSTs), focusing on the role of hydrogen gas bubbles and metallic protrusions within the dielectric oil.

Methodology:

The researchers employed a self-consistent plasma-fluid model using COMSOL Multiphysics and a custom-built solver (COPAIER) to simulate PD ignition. They considered various bubble sizes, positions, and the presence of protrusions under both high-frequency (50 kHz) and low-frequency (quasi-DC) voltage conditions.

Key Findings:

  • Larger bubbles significantly increase the probability and severity of PD events, leading to lower inception voltages.
  • Metallic protrusions in contact with bubbles drastically reduce the PD inception voltage, especially with sharper protrusions.
  • Even under typical operating conditions (66 kV RMS), bubbles larger than 2.5 mm can initiate PDs.
  • The bubble size effect is primarily attributed to charge gain/loss timescales rather than electrostatic field enhancement.

Main Conclusions:

Minimizing gas impurities and ensuring smooth winding surfaces are crucial for mitigating PD risk in SSTs. Even small bubbles can pose a hazard, particularly under transient overvoltage or in the presence of protrusions.

Significance:

This research provides valuable insights for designing reliable high-voltage transformers, highlighting the importance of considering even seemingly minor defects like gas bubbles and surface irregularities.

Limitations and Future Research:

The study assumes immobile, spherical bubbles. Future research could incorporate bubble deformation, movement, and interaction with fluid flow for a more comprehensive analysis. Additionally, coupling plasma dynamics with fluid and thermo-dynamics could provide a more realistic representation of the transformer environment.

edit_icon

Customize Summary

edit_icon

Rewrite with AI

edit_icon

Generate Citations

translate_icon

Translate Source

visual_icon

Generate MindMap

visit_icon

Visit Source

Stats
Typically, more than 70% of transformer faults are internal, initiated via partial discharges inside transformer insulation. Stray gassing in electrical insulating oils heated at relatively low temperatures favors the production of saturated carbon gas bubbles and especially hydrogen (H2) in temperatures below 120 °C. The primary winding in the simulation is fixed at position z=0, whilst the secondary winding is fixed at z = -50 mm (overall gap 50 mm) and is considered grounded (V0 = 0). The bubble is assumed spherical and the H2 plasma model is applied only inside the bubble. The gas bubble diameter varies between 1 and 4.5 mm. The bubble center is positioned at different coordinates in terms of x (radius r) and z, with fixed y = 0 (axis) and a sector angle 180° (half circle). The bubble is placed inside the dielectric fluid (insulation medium) and a fixed relative permittivity value ε = 3 is considered. The temperature is 360 K. For the high-frequency simulations, an AC applied voltage with a frequency (f) of 50 kHz is considered. For the low-frequency simulations, a DC voltage ramped up to its maximum value with a long rising time (~ 100 μs) is used. The study considers a 66 kV operating RMS voltage (peak value 93 kV) for the low-frequency simulations. The critical PD charge for PD inception is set to 10 pC.
Quotes
"The presence of bubbles in a dielectric fluid increases the breakdown probability." "The initial stage of the transformer oil breakdown is caused by a gas bubble formed by evaporation due to local heating in high electric field regions of the electrode surface." "As the PD is most probable when small amounts of gases are present, several models [8] have been developed to simulate the PD in voids (e.g., gas bubbles)."

Deeper Inquiries

How can machine learning be utilized to predict and prevent partial discharge occurrences in high-voltage transformers based on real-time monitoring data?

Machine learning (ML) presents a powerful tool for predicting and preventing partial discharge (PD) occurrences in high-voltage transformers by leveraging real-time monitoring data. Here's a breakdown of how ML can be implemented: 1. Data Acquisition and Preprocessing: Sensors: Deploy a network of sensors within the transformer to capture relevant parameters. These may include: Ultrasonic sensors: Detect the acoustic emissions produced by PD. High-frequency current transformers: Measure the transient currents associated with PD. Chemical sensors: Monitor the levels of gases (like hydrogen, methane, acetylene) generated by PD activity. Temperature sensors: Detect localized heating caused by PD. Data Collection: Gather real-time data streams from these sensors. Preprocessing: Cleanse the data to handle noise, outliers, and missing values. This may involve techniques like filtering, smoothing, and imputation. 2. Feature Engineering: Extract meaningful features from the preprocessed data that are highly correlated with PD activity. Examples include: Statistical features: Mean, standard deviation, skewness, and kurtosis of sensor readings. Frequency-domain features: Dominant frequencies and spectral energy distribution from frequency analysis of sensor signals. Time-frequency features: Wavelet transform coefficients to capture transient patterns in the data. Feature selection: Identify the most informative features that contribute significantly to PD prediction. 3. Model Training and Selection: Choose an appropriate ML model: Several ML algorithms can be employed, each with strengths and weaknesses: Supervised Learning: Support Vector Machines (SVMs): Effective for classifying PD patterns. Artificial Neural Networks (ANNs): Capable of modeling complex relationships between features and PD occurrence. Random Forests: Robust to noise and outliers, providing good generalization. Unsupervised Learning: Clustering algorithms (e.g., K-means, DBSCAN): Group similar PD patterns together, aiding in identifying different types of PD. Anomaly detection (e.g., One-Class SVM, Isolation Forest): Detect unusual patterns in the data that deviate from normal operating conditions, potentially indicating PD. Train the model: Use historical data with known PD events to train the selected ML model. The model learns the underlying patterns and relationships between the features and PD occurrences. Validation and Optimization: Fine-tune the model's parameters using a separate validation dataset to achieve optimal performance. 4. Real-time Prediction and Prevention: Deploy the trained model: Integrate the model into the transformer's monitoring system. Real-time monitoring: Continuously feed real-time sensor data into the model. PD prediction: The model predicts the likelihood of PD occurrence based on the input data. Preventive actions: If the model predicts a high risk of PD, trigger preventive measures such as: Alarms: Alert operators to potential issues. Voltage reduction: Temporarily lower the voltage to mitigate PD activity. Transformer shutdown: In severe cases, initiate a controlled shutdown to prevent catastrophic failure. Advantages of Using ML: Early detection: Identify PD at its initial stages, well before significant damage occurs. Improved accuracy: ML models can outperform traditional rule-based systems, especially in complex scenarios. Adaptive learning: Models can continuously learn from new data, improving their accuracy over time. Data-driven insights: Provide valuable information about the factors influencing PD, aiding in better transformer design and maintenance. Challenges: Data availability: Obtaining sufficient high-quality data for training and validation can be challenging. Model interpretability: Understanding the reasoning behind ML predictions can be difficult, especially with complex models. Real-time performance: Ensuring low latency and high throughput for real-time applications is crucial. By addressing these challenges and effectively implementing ML, we can significantly enhance the reliability and longevity of high-voltage transformers.

Could the presence of other gases besides hydrogen, such as nitrogen or oxygen, influence the partial discharge inception voltage differently?

Yes, the presence of gases other than hydrogen, particularly nitrogen and oxygen, can significantly influence the partial discharge (PD) inception voltage in transformer oil. Here's why: 1. Breakdown Strength: Different gases have different dielectric strengths (breakdown voltages). Nitrogen (N2): Generally has a higher breakdown strength than air (which is primarily nitrogen and oxygen) and significantly higher than hydrogen. Oxygen (O2): Has a lower breakdown strength compared to nitrogen. Hydrogen (H2): Has a much lower breakdown strength than both nitrogen and oxygen. Impact on Inception Voltage: Higher breakdown strength gases (like nitrogen) increase the PD inception voltage. This means a higher voltage is required to initiate a discharge. Lower breakdown strength gases (like hydrogen) decrease the PD inception voltage. Discharges are more likely to occur at lower voltages. 2. Chemical Interactions: Oxygen: Highly reactive: Oxygen can react with the transformer oil and other insulating materials, leading to: Oxidation: Degradation of the oil, reducing its dielectric strength and accelerating aging. Formation of acids and sludge: These byproducts can further contaminate the oil and promote PD. Nitrogen: Relatively inert: Nitrogen is less reactive than oxygen and is often used in transformers to create an inert atmosphere, reducing oxidation and PD. Hydrogen: Can indicate other problems: While hydrogen itself doesn't directly react with the oil as aggressively as oxygen, its presence often signals more serious issues like overheating or arcing, which can lead to further gas production and a cascade of problems. 3. Gas Mixtures: Real-world scenarios often involve gas mixtures. The composition of these mixtures will impact the overall breakdown strength and PD behavior. Synergistic effects: The presence of even small amounts of a gas with low breakdown strength (like hydrogen) in a mixture can significantly reduce the overall inception voltage compared to pure nitrogen or air. Example: Imagine a transformer with a small leak. Air (containing oxygen) enters and mixes with the oil. The oxygen reacts with the oil, degrading it and producing small amounts of hydrogen. This gas mixture, even with a small amount of hydrogen, will have a lower PD inception voltage than the original oil with only dissolved nitrogen. In summary: The presence of gases other than hydrogen, especially nitrogen and oxygen, plays a crucial role in determining the PD inception voltage in transformers. Gases with higher breakdown strengths generally increase the inception voltage, while those with lower breakdown strengths decrease it. Chemical interactions, particularly with oxygen, can further degrade the oil and promote PD. The composition of gas mixtures is critical, as even small amounts of low-breakdown-strength gases can significantly impact the overall inception voltage.

If we consider the transformer as a complex system, how can we apply principles of chaos theory to understand the seemingly unpredictable nature of partial discharge events?

While partial discharge (PD) events in transformers might appear random, applying principles of chaos theory can provide valuable insights into their underlying dynamics and potentially improve prediction capabilities. Here's how: 1. Transformers as Complex Systems: Interconnected components: Transformers consist of numerous interconnected components (windings, insulation, core, cooling system) with complex interactions. Nonlinear behavior: The electrical and thermal properties of these components often exhibit nonlinear behavior, meaning small changes in input can lead to disproportionately large effects. Sensitivity to initial conditions: Slight variations in manufacturing, operating conditions (temperature, load, voltage fluctuations), and the presence of defects can significantly influence PD behavior. This sensitivity is a hallmark of chaotic systems. 2. Applying Chaos Theory Principles: Phase Space Reconstruction: Representing the system's state: Instead of analyzing individual PD events in isolation, chaos theory suggests reconstructing the transformer's "phase space." This involves using time-delayed measurements of a single variable (e.g., PD magnitude, acoustic emissions) to create a multi-dimensional representation of the system's state. Identifying patterns: By analyzing trajectories in this phase space, we can potentially uncover hidden patterns and attractors that govern PD dynamics. Lyapunov Exponents: Quantifying sensitivity: Lyapunov exponents measure the rate of divergence or convergence of nearby trajectories in phase space. Positive Lyapunov exponents indicate chaotic behavior and sensitivity to initial conditions. Predictability limits: Calculating Lyapunov exponents can help estimate the time horizon over which PD predictions might be reliable. Fractal Analysis: Characterizing complexity: PD patterns often exhibit self-similarity across different scales, a characteristic of fractals. Fractal analysis can quantify this complexity and potentially distinguish between different types of PD. 3. Benefits and Challenges: Benefits: Enhanced understanding: Chaos theory can provide a deeper understanding of the complex interplay between various factors influencing PD. Improved prediction: By identifying underlying patterns and predictability limits, we can potentially develop more accurate PD prediction models. Targeted mitigation: Understanding the chaotic nature of PD can guide the development of more effective mitigation strategies, focusing on controlling key parameters and influencing the system's trajectory. Challenges: Data requirements: Chaos theory analysis often requires large amounts of high-resolution data, which might be challenging to obtain from transformers. Computational complexity: Reconstructing phase space and calculating chaos theory metrics can be computationally intensive. Interpretation: Translating chaos theory findings into practical insights and actionable recommendations for transformer operation and maintenance requires careful interpretation. In conclusion: While chaos theory presents challenges, its application to understanding PD in transformers holds significant promise. By embracing the complex and nonlinear nature of these systems, we can potentially move beyond simplistic models and develop more effective strategies for predicting, preventing, and mitigating the detrimental effects of partial discharge.
0
star