Machine Learning-Based State Estimation for an Actual Transmission System Using Field PMU Data

Time-synchronized state estimation (SE) is vital for ensuring real-time situational awareness in modern power systems; however, full system observability using phasor measurement units (PMUs) often faces challenges due to cost and deployment issues. This study investigates the application of deep neural networks for effective PMU-timescale state estimation in real-world scenarios where some transmission systems are not fully observable using actual data from a US power utility. We assess critical factors such as estimation accuracy, scalability, and computational performance under realistic operational conditions, providing insights into the feasibility of deploying advanced machine learning techniques for state estimation. Our findings indicate that neural network-based SE can enhance the reliability and efficiency of power system management, addressing the strict latency and robustness requirements in PMU-based frameworks despite existing unobservability constraints.