Abstract
Modern power systems are increasingly complex, interconnected, and reliant on real-time monitoring for reliable and secure operation. Accurate power system state estimation (PSSE) is essential for grid control, stability assessment, and decision-making. However, traditional PSSE methods particularly the widely used weighted least squares (WLS) estimator are highly sensitive to bad data and parameter inaccuracies, such as incorrect line impedances or admittances. In large-scale transmission systems, where cyber-physical uncertainties and data corruption are inevitable, this lack of robustness can lead to erroneous estimates, system instability, and even cascading failures.
This dissertation addresses the fundamental challenge of robust and scalable alternative current (AC) state estimation in transmission systems, with a particular focus on the impact of leverage points and parameter errors.
The primary objective of this work is to develop and validate a robust distributed PSSE algorithm that maximizes the breakdown point while maintaining statistical efficiency and scalability. The proposed approach builds upon robust statistical theory and more specifically the so-called robust S-estimators, which offer high breakdown points, and extends them to a distributed optimization framework suitable for large-scale trans- mission networks.
The dissertation is addressing different measurements, bad data, and modeling configurations. The estimator is applied to a system with hybrid measurements, combining traditional SCADA and PMU data. Then, the focus is narrowed to purely linear measurements, allowing for analytical insights into estimator performance under simplified assumptions.
Extensive validation is carried out using the IEEE 14-bus and 30-bus test systems, under various configurations of bad data and parameter errors. The results demonstrate that the proposed robust S-based distributed estimator significantly outperforms other estimators such as WLS with bad data detection (BDD) method and the least absolute value (LAV) in terms of estimation accuracy, outlier resistance, and parameter errordetection. Furthermore, the algorithm exhibits promising scalability, with computation time and communication overhead remaining acceptable as system size increases, making it a promising candidate for real-time or near-real-time deployment in practical power system environments.