Machine Learning Methods for State Estimation in Sparse Distribution Networks

The widespread deployment of renewable energy sources and the increasing uptake of electric vehicles contribute to greater complexity in managing power flows within distribution grids. Understanding system dynamics and implementing control actions are crucial for effective grid management. While smart meters, accurate sensors, communication networks facilitate measurements at specific voltage nodes, obtaining measurements at all nodes across a network is costly and, in some cases, impractical. In response to these challenges, this paper firstly compares different state estimation methods based on available measurements and secondly, explains why an artificial neural network-based state estimation method is the most effective. The proposed method utilizes data-driven techniques to deliver precise estimations of voltages and currents within distribution networks, offering valuable insights into flexibility management and control processes. This approach also considers the phase angle effect of voltage and holds significant promise in improving awareness of the situation for power system operators, enhancing their understanding of evolving grid dynamics. The robustness and effectiveness of this method are validated through missing data cases, demonstrating its adaptability to diverse operational conditions.