Bayesian-optimized LSTM-DWT approach for reliable fault detection in MMC-based HVDC systems

Area of Science:

• Electrical Engineering
• Power Systems
• Renewable Energy Integration

Background:

• Europe's increasing reliance on offshore wind power necessitates robust long-distance High Voltage Direct Current (HVDC) transmission.
• Modular Multi-Level Converters (MMCs) are crucial for modern HVDC systems but face challenges with rapid DC fault current increases.
• Existing fault detection methods often struggle with accuracy and require manual threshold adjustments.

Purpose of the Study:

• To develop a novel, highly accurate fault identification and classification method for HVDC transmission lines.
• To address the challenge of rapid DC fault current rise in MMC-based HVDC systems.
• To enhance the efficiency and robustness of DC grid protection against various faults and disturbances.

Main Methods:

• Integration of Long Short-Term Memory (LSTM) networks with Discrete Wavelet Transform (DWT) for feature extraction and fault classification.
• Development of a three-level relay system with multiple time windows (1 ms, 1.5 ms, 2 ms) for precise fault detection over extended distances.
• Application of Bayesian Optimization for efficient hyperparameter tuning of the LSTM model.

Main Results:

• The proposed LSTM-DWT framework achieved an average recognition accuracy rate of 99.04% across diverse testing scenarios.
• The system demonstrated 100% resilience against external faults and disturbances, proving its robustness.
• The algorithm successfully detected faults up to 480 ohms, outperforming traditional methods reliant on multiple manual thresholds.

Conclusions:

• The novel LSTM-DWT based fault detection system offers a highly accurate and robust solution for protecting HVDC transmission lines.
• This intelligent, single-model approach enhances DC grid protection efficiency and reliability, particularly in integrated renewable energy systems.
• The proposed method provides a significant advancement over traditional fault detection schemes, offering superior performance and adaptability.