Surrogate Modeling & PINNs for OPF
Machine learning and physics-informed neural network surrogates for Optimal Power Flow (Sustainability, 2024).
Solving AC and DC Optimal Power Flow (OPF) repeatedly in real-time grid operations is computationally bottlenecked by non-convex physical constraints and high renewable penetration.
This research investigates machine learning surrogate models and label-free Physics-Informed Neural Networks (PINNs) that embed power balance and physical network constraints directly into the training loss, achieving fast, high-accuracy OPF dispatch solutions compared to analytical solvers.
Publication:
- S. Mohammadi, V. Hai-Bui, W. Su, B. Wang (2024). “Surrogate Modeling for Solving OPF: A Review,” Sustainability, vol. 16, no. 22, p. 9851. DOI: 10.3390/su16229851