Enhanced physics-informed neural networks (PINNs) for high-order power grid dynamics

TL;DR

This work tackles the challenge of accelerating high-fidelity power-grid dynamic simulations in renewables-rich networks by developing enhanced physics-informed neural networks (PINNs) for high-order, high-dimensional ODEs that describe synchronous generators (SGs) and grid-following inverters (GFLs). The authors introduce four key enhancements—architecture/hyperparameter tuning, multiobjective loss normalization via Utopia/Nadir points, adaptive gradient balancing, and sequence-to-sequence learning—and apply them to a nonlinear fourth-order SG model and a 17-state inverter model. Preliminary results show that sgPINN can accurately predict SG dynamics, while invPINN remains more challenging due to faster, stiffer inverter dynamics, motivating further development including supervised training and parameter estimation with PMU data. Overall, the enhanced PINN framework has potential to enable faster, reliable transient stability analyses essential for planning and operating a renewables-dominated grid.

Abstract

We develop improved physics-informed neural networks (PINNs) for high-order and high-dimensional power system models described by nonlinear ordinary differential equations. We propose some novel enhancements to improve PINN training and accuracy and also implement several other recently proposed ideas from the literature. We successfully apply these to study the transient dynamics of synchronous generators. We also make progress towards applying PINNs to advanced inverter models. Such enhanced PINNs can allow us to accelerate high-fidelity simulations needed to ensure a stable and reliable renewables-rich future grid.

Figures (2)

• Figure 1: Comparison of PINN predictions and true solutions for one test initial condition.
• Figure 2: Inverter simulation results highlighting faster dynamics that are more challenging to capture, than with SGs.