Advances in decentralized state estimation for power systems

Distributed learning via network diffusion is a popular trend in signal processing, which addresses the need of obtaining scalable analytics from networked sensor systems. This paper describes relevant advances in distributed power system state estimation (PSSE) via diffusion. Considering a hybrid sensor measurements system, we show that the Gauss-Newton approach, typically favored in PSSE, can be used as a primitive to derive a gossip-based algorithm that outperforms first order diffusion methods proposed in the literature. We also study analytically and numerically the dependency between measurement placement, grid topology and physical parameters, communication network and the performance of the decentralized PSSE.