Neuro-Physics Hybrid State Estimation of Distribution System With Smart Meter Voltage Measurements

The lack of measurements poses a significant challenge for the medium-voltage distribution system state estimation (MV-DSSE). The proliferation of smart meters can alleviate this issue by providing abundant measurements at customer-end points in low-voltage (LV) networks; however, their voltage measurements cannot be directly incorporated into the MV-DSSE. This paper proposes a novel neuro-physics hybrid SE model for MV networks, incorporating smart meter voltage measurements from LV networks into MV-DSSE. In our approach, the physical SE model is adopted for the MV network, while a neural network (NN)-based model is trained to represent the relationship between the MV state variables and the LV network voltage measurements. The hybrid SE model does not require the LV network model, which is often unavailable to utility operators. The test results demonstrate that the proposed model enhances the accuracy and robustness of MV-DSSE by incorporating smart meter voltage measurements in LV networks.