Physics-Informed Learning of Proprietary Inverter Models for Grid Dynamic Studies

2025-07-21

AI Summary

To address insufficient dynamic simulation accuracy in power grids caused by the opacity of inverter internal control parameters, this paper proposes a Physics-Informed Latent Neural ODE (PI-LNM) model. The method embeds physical constraints—such as power conservation and structural forms of dynamical equations—into a neural ordinary differential equation framework, integrating latent-variable modeling with end-to-end learning to achieve high-fidelity representation of grid-forming inverter dynamics. Compared to purely data-driven approaches, PI-LNM significantly improves prediction accuracy of transient responses while preserving model interpretability, thereby enabling robust stability analysis and control parameter tuning. This work establishes a novel paradigm for physics-informed, data-driven modeling of black-box power electronic devices, bridging mechanistic understanding and empirical learning in modern power system simulation.

Technology Category

• Machine Learning: Imitation Learning & Inverse Reinforcement Learning
• Search and Optimization: Sampling/Simulation-based Search
• Cognitive Modeling & Cognitive Systems: Conceptual Inference and Reasoning

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Systems and Infrastructure for Web, Mobile and WoT: Energy management for devices in mobile Web and WoT environments
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs

Abstract

This letter develops a novel physics-informed neural ordinary differential equations-based framework to emulate the proprietary dynamics of the inverters -- essential for improved accuracy in grid dynamic simulations. In current industry practice, the original equipment manufacturers (OEMs) often do not disclose the exact internal controls and parameters of the inverters, posing significant challenges in performing accurate dynamic simulations and other relevant studies, such as gain tunings for stability analysis and controls. To address this, we propose a Physics-Informed Latent Neural ODE Model (PI-LNM) that integrates system physics with neural learning layers to capture the unmodeled behaviors of proprietary units. The proposed method is validated using a grid-forming inverter (GFM) case study, demonstrating improved dynamic simulation accuracy over approaches that rely solely on data-driven learning without physics-based guidance.

Problem

• Emulate proprietary inverter dynamics for grid simulations
• Address lack of disclosed internal controls from OEMs
• Improve accuracy over purely data-driven learning methods

Innovation

• Physics-informed neural ODEs for inverter dynamics
• Integrates system physics with neural learning layers
• Improved accuracy over data-driven-only approaches