Distributed Fault Detection for Cyber-Physical Systems With Application to Power Network System.

Area of Science:

• Cyber-Physical Systems (CPS)
• Control Systems Engineering
• Distributed Signal Processing

Background:

• Cyber-physical systems (CPS) are complex, comprising numerous interconnected subsystems.
• Accurate fault detection is crucial for the reliable operation of CPS.
• Traditional centralized fault detection methods face scalability and communication challenges in large-scale CPS.

Purpose of the Study:

• To develop a distributed fault detection scheme for a class of CPS.
• To enhance the robustness and accuracy of fault detection in individual subsystems.
• To ensure fault detection operates without relying on global information exchange.

Main Methods:

• State estimation for each subsystem using a distributed Kalman filter (DKF), incorporating neighbor information.
• Design of local residual generators leveraging the DKF for each subsystem.
• Parameter tuning of the DKF to minimize estimation error and covariance bounds for residual robustness.
• Utilization of instantaneous and sliding window T² test statistics for residual evaluation and threshold setting.

Main Results:

• A distributed fault detection scheme where each subsystem communicates only with its neighbors.
• Demonstrated robustness of the residual signals through optimized DKF parameters.
• Established fault detection thresholds based on T² test statistics.
• Provided a condition guaranteeing mean square boundedness of estimation error in fault-free scenarios.

Conclusions:

• The proposed distributed fault detection scheme effectively enables local fault detection in CPS subsystems.
• The method ensures robustness and reliability by minimizing estimation errors and utilizing neighbor information.
• The scheme's distributed nature enhances scalability and reduces communication overhead in complex CPS.