A Study on Super-Resolution Voltage Prediction in Distribution System Operator based on Spatiotemporal Graph Learning

Abstract

The increasing penetration of distributed energy resources (DERs) in distribution network the need for high-precision monitoring by Distribution System Operator (DSO) is growing rapidly. Unlike transmission systems, distribution network contain a significantly larger number of buses, which poses economic and computational challenges in acquiring, storing, and managing high-resolution measurement data. Consequently, even when high-resolution data collection is technically feasible at the local level, DSOs typically receive and utilize data at low-resolution intervals due to these limitations. This constraint accurate capture of DER generation and load fluctuations between monitoring cycles, potentially leading to system instability. This study proposes a novel two-stage framework for super-resolution voltage prediction in distribution network, combining Gaussian Process Regression (GPR) for high-resolution load estimation with uncertainty quantification, and a hybrid Graph Convolutional Network–Long Short-Term Memory (GCN-LSTM) model to capture spatio-temporal characteristics of the grid. To evaluate the performance of the proposed method, case studies were conducted using the Modified IEEE 13-Bus System. The voltage prediction results demonstrate that the proposed method outperforms conventional model. In conclusion, this research provides a practical solution to emerging challenges in distribution network operation due to the growing deployment of DERs.