Anomaly detection method for power system information based on multimodal data.

Area of Science:

• Electrical Engineering
• Control Systems
• Data Science

Background:

• Modern power systems face increasing complexity, necessitating robust anomaly detection for operational security.
• Conventional methods using single-domain data struggle to capture the full dynamic behavior of power systems.
• Limitations in traditional techniques hinder comprehensive analysis and effective threat identification.

Purpose of the Study:

• To introduce a novel multimodal approach for enhanced anomaly detection in power systems.
• To improve detection accuracy and robustness by integrating diverse data domains.
• To provide a versatile framework for securing critical infrastructures.

Main Methods:

• Integration of time-domain and frequency-domain data for anomaly detection.
• Leveraging multimodal data to capture both temporal patterns and spectral signatures.
• Comparative analysis against conventional single-domain techniques.

Main Results:

• Achieved a detection accuracy of 97.6%, significantly outperforming baseline methods.
• Demonstrated enhanced robustness and comprehensive analysis of system behavior.
• Validated the effectiveness of the multimodal approach in complex power systems.

Conclusions:

• The multimodal approach offers superior performance for power system anomaly detection.
• Early identification of security threats and improved system reliability are key benefits.
• This framework provides a versatile solution for anomaly detection in critical infrastructures.