Monitoring of low-voltage grids using artificial neural networks and its field test application based on the beeDIP-platform

Abstract

Cost-effective operation of low-voltage (LV) grids with the increasing share of distributed generators, electric vehicles, and associated volatility and unpredictability requires novel operational management strategies from the distribution system operators. To reduce the cost for grid reinforcements, monitoring of the distribution grids is of crucial importance. However, reliable measurement of the entire LV grid is economically rather impractical since it would require a very high density of measuring instruments, of which there are only very few in LV grids today. Thus, we need suitable grid monitoring methods that provide sufficiently accurate results even with a relatively low density of direct measurements. For this paper, we applied artificial neural networks (ANN)-based grid state estimation method, which has already been successfully used for medium-voltage grid monitoring, to LV grids. The performance of the ANN-based method for different measurement configurations is investigated in simulations with SimBench LV grids and is currently researched in a field test based on the beeDIP-Platform within a German suburban grid.