Ideal sensor placement and state estimation in distribution grids

Low-voltage (LV) distribution grids commonly suffer from a low level of observability due to the lack of measurement devices in the grid. This impedes a reliable state estimation (SE) in such grids, which is the foundation for further intelligent control algorithms. Furthermore, the increasing penetration of intermittent and volatile, yet large, loads and generators such as photo-voltaic installations or electric vehicles, pose a risk for LV distribution grids. In order to develop more resilient LV distribution grids, observability is therefore key. In this paper a novel algorithm to determine ideal sensor locations in distribution grids is presented. The basis of this novel algorithm are large Monte-Carlo-simulated datasets of power flows in LV grids. The algorithm then employs principal component analysis and other statistical measures to determine the ideal placement of sensors within LV grids. In combination with an established machine learning-based SE tool, this leads to fast and reliable SE for which minimal measurement resources are required.