Error propagation in the state estimation and prediction of distribution grids with a limited database

To evaluate this question under realistic conditions, a dataset is created that combines realworld MV data with synthetic, yet physically consistent and representative, LV data. LVSE is implemented as a weighted least squares estimator to infer bus voltages and line loadings. In addition, the paper assesses low-voltage state forecasting (LVSF) under the same measurement constraints using a data-driven deep neural network.

Results show that LVSE can estimate both bus voltages and line loadings with sufficient accuracy even when neither transformer nor feeder measurements are available. However, this requires at least 20% penetration of intelligent metering systems (iMSys) in the LV network combined with MVSE-based transformer power and voltage estimates.

For forecasting, the results indicate that with 20% iMSys penetration and high-quality MVSE inputs, LVSF achieves feasible accuracy for a 1-hour prediction horizon. For longer horizons, forecast errors increase substantially, and 2-hour-ahead states are not consistently predicted with sufficient accuracy. Overall, the findings suggest that digitizing secondary substations solely for LVSE and LVSF purposes can be avoided when iMSys coverage is adequate and

MVSE is available. Thus, metering infrastructure and operational costs can be reduced while secure LV operation is maintained.