State Estimation for Power Distribution System Using Graph Neural Networks

State estimation is critical to maintaining system stability and reliability as it enables real-time monitoring of the power system operation and facilitates fault detection, minimizing the risk of power outages and improving overall system performance. This paper presents a state estimation method based on graph neural networks, aiming to improve time efficiency and extended observability. Graph neural networks can aggregate information and dependencies from voltage and power measurement at the critical buses, making them more effective for state estimation on non-grid structured data. The IEEE 123-bus system is used as a case study to evaluate comprehensively the state estimation performance. The proposed model provides a better performance for mapping measurement data with states compared to other neural networks.