Data‐driven sensor placement for state reconstruction via POD analysis

Abstract

In this study, a data-driven framework to determine the optimal sensor locations for power system oscillation monitoring and state reconstruction is proposed. In this approach, candidate sensor locations are selected sequentially using proper orthogonal decomposition (POD) analysis of a reduced-order binary measurement matrix. Using criteria inferred from the spatial structure of the POD modes, sensor locations having large magnitudes and small cross couplings between modes at the candidate locations are selected from the measurement matrix. The method can be used to reconstruct global system dynamics using a few sensor measurements as well as for the prognosis of power system oscillatory behaviour. The developed procedures are applied to the IEEE 39-bus test system and to a realistic 5449-bus model of a large power system.