GridSFM: A Foundation Model for Solving AC Optimal Power Flow
GridSFM, a 15-million-parameter graph model, solves AC optimal power flow on grids up to 10,000 buses.
GridSFM is a 15-million-parameter physics-inspired graph neural network pretrained across 54 grid topologies of 500 to 4,000 buses for AC optimal power flow. It reports a 2.45% zero-shot generation-cost error on held-out operating conditions for a 10,000-bus case, with no degradation as system size grows. Physics-informed fine-tuning based on Newton's method adapts it to unseen grids up to 10,000 buses using 100 solved instances and beats more heavily trained single-topology models as a warm start. The authors release models, data, and code and prove an elastic relaxation whose feasible set is contractible.
- 15M-parameter GNN pretrained on 54 topologies of 500 to 4,000 buses.
- Zero-shot generation-cost error was 2.45% on a 10,000-bus case.
- Newton-based fine-tuning adapts to new grids with 100 solved instances.
- Authors release models, data, and code and prove an elastic relaxation.
Full article240 words · extracted from arxiv.org · click to collapse
We introduce GridSFM, a framework that combines a pretrained foundation model across grid topologies with physics-informed fine-tuning for solving AC Optimal Power Flow (AC-OPF) at scale. It is a $15$ million parameter physics-inspired graph neural network pretrained across $54$ topologies of $500$ to $4{,}000$ buses. Our model attains a $2.45\%$ zero-shot generation-cost error on a $10{,}000$ bus case held-out operating conditions with no degradation as system size grows. Building on this, we pair the pretrained backbone with a physics-informed fine-tuning design based on Newton's method for power flow. With only $100$ solved instances, GridSFM adapts to unseen grids up to $10{,}000$ buses. We show it out performs single topology, dedicated neural network models that are trained more data, both in terms of cost and solver iterations when deployed as warm starting points. In designing this foundation model, we overcome the fact that the feasible set for AC-OPF can be disconnected. This is an obstruction that prevents any continuous neural network from approximating the solution map. To do so, we lift the problem and relax its constraints with logarithmically penalized slacks. We prove that the resulting elastic feasible set is contractible, that the AC-OPF minimizers remain minimizers of the elastic problem above an explicit penalty threshold, and that projecting an approximate solution back onto the AC-OPF feasible set is well posed. We release all models, data, and code so that the community can build on a shared starting point for AC-OPF.
Text extracted automatically; images, tables and formatting may be missing. Original: https://arxiv.org/abs/2609.30173