Dynamic load altering attack detection in cyber physical power system based on improved Kalman/H_∞ co-filtering.

Area of Science:

• Cyber Physical Power Systems (CPPSs)
• Control Systems Engineering
• Network Security

Background:

• Cyber Physical Power Systems (CPPSs) are vulnerable to Dynamic Load Altering Attacks (DLAA).
• Accurate state estimation and robust attack detection are crucial for CPPSs' stability and security.
• Existing methods struggle with unknown-statistics noise and non-Gaussian disturbances.

Purpose of the Study:

• To propose an advanced attack detection scheme for closed-loop DLAA in CPPSs.
• To enhance the accuracy of state estimation and attack detection capabilities.
• To address challenges posed by DLAA and noise disturbances in CPPSs.

Main Methods:

• Developed a discrete-time CPPS model incorporating DLAA and unknown-statistics noise.
• Proposed an improved Kalman/H∞ co-filter for state estimation, utilizing MFAKF for Gaussian noise and H∞ filter for non-Gaussian disturbances.
• Designed a cosine similarity matching algorithm for anomaly detection based on state estimation deviations.

Main Results:

• The proposed MFAKF-HF significantly reduced RMSE by 75% compared to MFAKF and 62% compared to H∞ filtering for state ω1 in the IEEE 3-machine 6-bus system.
• Demonstrated improved accuracy in state estimation under DLAA and noise conditions.
• Validated the robustness and effectiveness of the integrated estimation and detection scheme.

Conclusions:

• The developed scheme offers superior performance in state estimation and DLAA detection within CPPSs.
• The combination of MFAKF and H∞ filtering effectively handles diverse noise characteristics.
• The cosine similarity matching provides a reliable method for identifying cyberattacks in power systems.