.png){kind=icon}
.png){kind=icon}
Return
Modified hard-constrained PINNs for physically consistent microgrid transient response modeling
.png){kind=icon}
PRE
AI
DOI：10.1016/j.ijepes.2025.110726.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Accurate dynamic equivalence modeling of microgrids is critical for distribution system operators, yet existing methods face trade-offs between physical consistency and scalability in high-dimensional parameter spaces. This work proposes a physically consistent parameter tuning paradigm for microgrid transient equivalence modeling, where the grey-box model structure serves as prior physical knowledge, and the hard-constrained PINN enables automated discovery of high-dimensional parameters under nonlinear dynamics. Compared to conventional 'structure-first' grey-box model for transient response analysis, our approach unifies physics fidelity with data-driven adaptability, achieves a new balance between interpretability and generalization in microgrid equivalent modeling. The key innovation lies in the implementation of physically consistent parameter tuning for microgrid transient analysis, where the hard-constrained PINN acts as a differentiable simulator to navigate high-dimensional spaces while strictly adhering to dynamic ordinary differential equations (ODEs). The paradigm eliminates conventional manual tuning problems through gradient-aware exploration of non-convex landscapes, resolving the trade-off between fidelity and measurement noise adaptation. Experimental validation on a real-time digital simulation (RTDS) control-hardware-in-loop (CHIL) platform demonstrates that the PINN-based parameters improve dynamic response accuracy and reduce error margins compared to traditional methods.
Keywords:
Physics-informed neural network
Microgrid
Equivalent model
Parameters estimation
Ordinary differential equation
Journal
I
IF:
5
Papers:
1.1W
Citations:
3.1W