State Estimation in Partially Observable Power Systems via Graph Signal Processing Tools

Area of Science:

• Electrical Engineering
• Computer Science
• Applied Mathematics

Background:

• Power system state estimation is crucial for grid operation and control.
• Conventional methods struggle with unobservable systems due to insufficient measurements.
• Existing solutions often rely on inaccurate pseudo-data or unavailable large datasets.

Purpose of the Study:

• To address the challenge of state estimation in power systems with limited measurements.
• To develop novel methods overcoming information deficits in unobservable grids.
• To propose an optimized sensor placement strategy for improved estimation.

Main Methods:

• Validated the graph smoothness property of power system states (voltages) theoretically and empirically.
• Developed a regularized graph signal processing weighted least squares (GSP-WLS) state estimator leveraging state smoothness.
• Introduced a sensor placement strategy tailored for the GSP-WLS estimator's performance optimization.

Main Results:

• GSP methods reduced estimation error magnitude by up to two orders compared to existing approaches.
• Achieved these results using only 70 sampled buses in the IEEE 118-bus system.
• Improved bad data detection probability by up to 30% for the same false alarm rate in unobservable systems.

Conclusions:

• Proposed GSP methods enable accurate state estimation in unobservable power systems.
• Significantly reduce the number of required measurement sensors.
• Offer a robust solution for power system monitoring and control under data scarcity.