Application of nonlinear Kalman filters to the identification of customer phase connection in distribution grids

Phase topology identification in low-voltage distribution networks: A Bayesian approach

“…In [24] the same authors extended this idea but considered the existence of photovoltaic panels. An approach to identifying low-voltage customers using Kalman filters was proposed in [25]. The authors of this previous paper established three Kalman filters, one for each phase, in which the state variables represent the phase connection of the customer, and the measurement of the filters is the energy delivered by the secondary substation.…”

“…However, these test feeders mostly represent North American distribution networks, which have a different distribution schema than the European ones. As previously described, the North American energy distribution is mostly done in medium voltage, in which many singlephase transformers are used to supply a single customer or a small group of customers [25,35]. This distribution schema contrasts with the European model, in which the secondary distribution (the last mile) is mainly done with a 4-wire low-voltage network (400/220 V) provided by a delta-wye power transformer with a grounded neutral connected to the medium-voltage network [6].…”

“…The grid was composed of 8087 loads distributed among 30 secondary distribution substations. This network has also been used by other authors to test phase identification methods [25].…”

Phase Identification of Distribution System Users Through a MILP Extension of State Estimation

“…[13] uses principal component analysis and conservation of energy constraints. In [14], a method is proposed that relies on hourly energy data and nonlinear Kalman filters. Relying on power/energy measurements like [9]- [14] is convenient, as these are more often available than voltage measurements, as they are also used for billing purposes.…”

“…In [14], a method is proposed that relies on hourly energy data and nonlinear Kalman filters. Relying on power/energy measurements like [9]- [14] is convenient, as these are more often available than voltage measurements, as they are also used for billing purposes. However, these methods can lead to inaccurate PI depending on the magnitude and pattern similarities of user demand, particularly in networks with many users.…”