Grid-forming Control of Converter Infinite Bus System: Modeling by Data-driven Methods

TL;DR

The paper addresses data-driven identification of grid-forming converter dynamics in an infinite-bus setup under disturbances. It compares Sparse Identification of Nonlinear Dynamics (SINDy) and Deep Symbolic Regression (DSR) on synthetic data generated for grid-forming control. DSR achieves near-perfect fidelity (R^2 ≈ 0.99) but with substantially higher computational cost (approximately 11×) than SINDy, while SINDy provides efficient, interpretable models with robust performance. The work suggests hybrid approaches to balance accuracy and runtime for real-time grid control in renewable-dominated power systems.

Abstract

This study explores data-driven modeling techniques to capture the dynamics of a grid-forming converter-based infinite bus system, critical for renewable-integrated power grids. Using sparse identification of nonlinear dynamics and deep symbolic regression, models were generated from synthetic data simulating key disturbances in active power, reactive power, and voltage references. Deep symbolic regression demonstrated more accuracy in capturing complex system dynamics, though it required substantially more computational time than sparse identification of nonlinear dynamics. These findings suggest that while deep symbolic regression offers high fidelity, sparse identification of nonlinear dynamics provides a more computationally efficient approach, balancing accuracy and runtime for real-time grid applications.

Figures (3)

• Figure 1: One-line diagram of the converter infinite bus system under grid-forming control mode.
• Figure 2: Identification of dynamic response of the LCL filter in a converter infinite bus system operating under grid-forming control mode under three sequential disturbances, using SINDy and DSR methodologies. The disturbances were introduced at different times: a change in active power reference $p^{\text{ref}}$ to 0.7 p.u. at $t = 0.5$ s, a change in reactive power reference $q^{\text{ref}}$ to 0.2 p.u. at $t = 1.0$ s, and a change in voltage reference $v^{\text{ref}}$ to 0.9 p.u. at $t = 1.5$ s. The subplots illustrate the model’s ability to capture the dynamics of various state derivatives under these disturbances; (a)$\frac{d}{dt}{i}_{r}^{\textnormal{cv}}$, (b)$\frac{d}{dt}{i}_{i}^{\textnormal{cv}}$, (c)$\frac{d}{dt}{v}_{r}^{\textnormal{filt}}$, (d)$\frac{d}{dt}{v}_{i}^{\textnormal{filt}}$, (e)$\frac{d}{dt}{i}_{r}^{\textnormal{filt}}$, and (f)$\frac{d}{dt}{i}_{i}^{\textnormal{filt}}$.
• Figure 3: Dynamic response characterization of the outer control loop in a converter infinite bus system operating under grid-forming control mode with three sequential disturbances, utilizing SINDy and DSR methodologies. The disturbances were introduced at different times: a change in active power reference $p^{\text{ref}}$ to 0.7 p.u. at $t = 0.5$ s, a change in reactive power reference $q^{\text{ref}}$ to 0.2 p.u. at $t = 1.0$ s, and a change in voltage reference $v^{\text{ref}}$ to 0.9 p.u. at $t = 1.5$ s. The subplots illustrate the model’s ability to capture the dynamics of various state derivatives under these disturbances; (a)$\frac{d}{dt}{\theta}^{\textnormal{oc}}$, (b)$\frac{d}{dt}{p}_{m}$, and (c)$\frac{d}{dt}{q}_{m}$.