Islanded microgrid security analysis method based on physical information neural network
By combining physical information neural networks and traditional power flow calculation models, the real-time performance and computational efficiency issues of static security analysis in islanded microgrids are solved, achieving efficient and adaptive security assessment, which is particularly suitable for high volatility scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- XIAN UNIV OF TECH
- Filing Date
- 2026-02-06
- Publication Date
- 2026-06-05
AI Technical Summary
Existing technologies lack a systematic methodology for static security analysis of isolated microgrids, making it difficult to meet real-time and computational efficiency requirements. Furthermore, data-driven methods are limited in application within power systems and cannot effectively utilize historical data to enhance analytical intelligence.
We employ a Physical Information Neural Network (PINN) approach, combining traditional power flow calculation models and data-driven methods. By constructing a deep learning model and embedding a physical constraint loss function, we achieve efficient and adaptive safety assessment.
It achieves efficient, adaptive and strongly constrained safety assessment of isolated microgrids, can dynamically modify the admittance matrix for N-1 verification, and combines computational efficiency with physical interpretability, making it suitable for high volatility scenarios.
Smart Images
Figure CN122159224A_ABST