Reconstruction of Low-Voltage Networks with Limited Observability

An electrical distribution network is the section of the power grid that links high voltage transmission lines to the end users. With an ever-increasing penetration of distributed energy resources and electric vehicles, the efficient management of distribution systems becomes more complicated and distribution system operators (DSOs) might encounter serious difficulties in guaranteeing the security of the network at Low-Voltage (LV) levels. To deal with this challenge, operators need to be able to assess preventively the system’s response to realistic scenarios of power consumption and production. This allows to implement effective preventive or corrective measures to avoid safety issues. This analysis can be performed through power flow studies [1], but reliable solutions require accurate information about the network topology and the physical characteristics of the lines. LV distribution networks are mostly operated radially. Their topology can change over time because of faults, maintenance and reconfiguration. At these voltage levels, most loads and appliances are connected to the network through a single phase and the neutral wire. These single-phase connections lead to a unbalanced system where voltage, current and power are different in all three phases, limiting the hosting capacity of the network. LV network operators do not always know how households, feeders and other appliances are interconnected. It can also happen that informations in their possession are not updated or correct. In other words, they usually lack of a reliable electrical model for the network. Such lack of knowledge hinders the efficient management and development of the system. The network topology identification is the mathematical process that allows to deduce this information. Effective identification methods for LV networks are thus essential for the development of smarter grids [2]. The goal of this work is to present a methodology that allows to infer the connections between nodes and to estimate line parameters of a LV network from time-series measurements provided by a limited amount of meters in the grid. An overview of existing methods is now presented, followed by a clear description of the network topology identification problem.