Automatic identification of low-voltage grids switches states using smart-meters data - Grenoble experiment

Low Voltage networks configuration can be adjusted to optimize assets, solve outages, or enhance power quality. These reconfigurations are achieved using switches, which shift a network branch from a substation or feeder to another. Yet, switch states are not monitored. Hence, each change needs to be followed by a manual update of the network Geographic Information System (GIS), which can be prone to errors. To tackle this issue, this paper proposes a method leveraging smart meters data to automatically identify switches states and correct the GIS. This method first decomposes the grid into smaller independent sub-grids. Then, for each sub-grid, the most probable opening point is identified using meters measurements. The method has been tested with GreenAlp, a DSO in the French Alps, over an area featuring 2,695 meters. Open switches could systematically be correctly identified except in specific cases (where there is not at least two meters on each side of the switch under consideration). Furthermore, the results did not suffer from the reduction of data depth to only one week, which is promising for its systematic use in operation.