Learning Power System Dynamics with Noisy Data Using Neural Ordinary Differential Equations
Shaorong Zhang, Koji Yamashita, Nanpeng Yu · 2024
The ability to learn complex system dynamics is crucial to enhancing the reliability and stability of power systems. In this paper, we develop a novel neural ordinary differential equation (ODE) based algorithm to predict the transient trajectories of power systems. To handle noisy measurements, we propose a noise-removal module, which is implemented before the neural ODE module. The proposed algorithm is validated using the IEEE 118-bus system. The numerical study results demonstrated the superior accuracy of the proposed model over the baseline neural network (NN) and its robustness against measurement noise. Furthermore, the analytics results verified the generalization performance across different fault durations and locations.