Acceleration of Power System Dynamic Simulations using a Deep Equilibrium Layer and Neural ODE Surrogate

TL;DR

A data-driven surrogate model based on implicit machine learning based on deep equilibrium layers and neural ordinary differential equations is proposed to learn a reduced order model of a portion of the full underlying system.

Abstract

The dominant paradigm for power system dynamic simulation is to build system-level simulations by combining physics-based models of individual components. The sheer size of the system along with the rapid integration of inverter-based resources exacerbates the computational burden of running time domain simulations. In this paper, we propose a data-driven surrogate model based on implicit machine learning -- specifically deep equilibrium layers and neural ordinary differential equations -- to learn a reduced order model of a portion of the full underlying system. The data-driven surrogate achieves similar accuracy and reduction in simulation time compared to a physics-based surrogate, without the constraint of requiring detailed knowledge of the underlying dynamic models. This work also establishes key requirements needed to integrate the surrogate into existing simulation workflows; the proposed surrogate is initialized to a steady state operating point that matches the power flow solution by design.

Figures (11)

• Figure 1: For large systems, modeling all devices in detail (above) becomes computationally intractable. This paper proposes a data-driven surrogate model that is designed to be integrated into simulations containing physics-based models (below).
• Figure 2: Types of applications of ML to power system dynamics.
• Figure 3: An overview of the proposed methodology. The different colors correspond to the focus of each subsection of the methodology.
• Figure 4: The proposed surrogate model. The blocks are color-coded according to their function---a key novelty lies in the combination of DEQ and NODE implicit layers with shared parameters.
• Figure 5: An expository result---(a) and (b) show the same quantities for the surrogate before and after training respectively. The left side shows the trajectories for the hidden states ($\boldsymbol{x}$) along with the predicted initial conditions ($\boldsymbol{\hat{x}}_0$) shown as black circles. The right side shows the model output (real and imaginary current).