Generative Adversarial Networks-Based Synthetic PMU Data Creation for Improved Event Classification

A two-stage machine learning-based approach for creating synthetic phasor measurement unit (PMU) data is proposed in this article. This leverages generative adversarial networks (GAN) generation and incorporates neural ordinary differential equation (Neural ODE) to guarantee underlying physical meaning. We utilize synthetically create massive eventful PMU data, which would otherwise be difficult obtain from the real world due critical energy infrastructure information (CEII) protection. To illustrate utility of such subsequent data-driven methods, we specifically demonstrate application using event classification by scaling up set. The addition a small set shown have improved accuracy 2 5 percent.