Abstract
The paper pioneers and optimizes a systematic approach to developing Physics-Informed neuro-Models (PIM) for the transient analysis of power grids interconnected with inverter-based resources. PIM serves as an effective online digital twin of power components, incorporating physical knowledge and preserving the system’s nonlinear differential structure while requiring only minimal data for training. Three contributions are presented: 1) An Physics Informed Neural Network (PINN)-enabled neuro-modeling approach for constructing an accurate ElectroMagnetic Transient (EMT) model; 2) A data-physics hybrid, multi-neural learning structure that demonstrates PIM’s adaptability at varying levels of data availability; 3) A balanced-adaptive PIM automatically optimizes the learning process while ensuring alignment with physical principles. Tests on rotating and static electrical components, as well as an IEEE test system, validate its efficacy for transient grid analysis under diverse operational scenarios.
| Original language | English |
|---|---|
| Pages (from-to) | 1806-1818 |
| Number of pages | 13 |
| Journal | IEEE Transactions on Consumer Electronics |
| Volume | 71 |
| Issue number | 1 |
| DOIs | |
| State | Published - 2025 |
Keywords
- EMT modeling
- data-physics hybrid learning
- inverter-based resources
- physics-informed machine learning
- self-adaptive learning
Fingerprint
Dive into the research topics of 'Neuro-Modeling Infused EMT Analytics'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver