Interpretable adaptive fault detection method for smart grid based on belief rule base.

Area of Science:

• Electrical Engineering
• Computer Science
• Artificial Intelligence

Background:

• Smart grid research prioritizes effective fault detection for system reliability and security.
• Traditional methods often sacrifice model interpretability for high accuracy, hindering operator trust.
• Complex models offer accuracy but lack transparency, posing challenges for understanding and validating results.

Purpose of the Study:

• To propose a new fault detection strategy that balances model interpretability and detection accuracy.
• To develop an adaptive interpretable belief rule base (AI-BRB) for smart grid fault detection.
• To enhance the trustworthiness and robustness of fault detection models in smart grids.

Main Methods:

• Developed an adaptive interpretable belief rule base (AI-BRB) for fault detection.
• Incorporated interpretability constraints into the model optimization process.
• Implemented adaptive updating of the search domain based on model accuracy to avoid local optima.

Main Results:

• The AI-BRB method achieved a balance between model interpretability and detection accuracy.
• Interpretability was maintained throughout the modeling, inference, and optimization stages.
• Adaptive search domain updates improved model robustness and prevented local optimal solutions.

Conclusions:

• The proposed AI-BRB method enhances both the accuracy and interpretability of smart grid fault detection.
• This approach offers a more transparent and trustworthy solution compared to existing methods.
• The AI-BRB strategy contributes to more reliable and secure smart grid operations.