Full Observability-Constrained PMU Placement Optimization Through State Estimation Accuracy Analysis

Power-system state estimation (SE) infers the most likely operating state of an electrical network from redundant, noisy measurements and the network model. This thesis investigates the impact of Phasor Measurement Units (PMUs) on SE accuracy and develops a methodology for selecting a practically superior PMU configuration when multiple placement solutions satisfy the same observability requirements. The Optimal PMU Placement Problem is formulated as an Integer Linear Programming (ILP) model that guarantees full network observability while minimizing the number of installed PMUs. The formulation is evaluated both with and without incorporating zero-injection buses, which enhance observability through Kirchhoff’s Current Law constraints. While the Optimal PMU Placement Problem often yields several solutions with the same minimum PMU count, these alternatives can lead to different estimation performance once measurement noise is considered. To resolve this ambiguity, a performance-driven selection framework is proposed. All minimum-PMU observable configurations (or a representative sample when their number is very large) are assessed using Monte Carlo simulations of weighted least squares state estimation. Configurations are ranked using mean absolute error (MAE). The approach is validated on IEEE 9-, 14-, 30-, 57-, and 118-bus test systems under defined noise models. Results show that observability-optimal placements are not necessarily equivalent in estimation quality, and that MAE-based ranking enables identifying a practically preferable PMU placement among multiple optimal solutions.