LUMINA: Foundation Models for Topology Transferable ACOPF

2026-03-04

AI Summary

This work addresses the challenge of developing scientific foundation models for power system optimization that simultaneously achieve strong generalization and strict adherence to physical constraints. To this end, the authors propose LUMINA, a framework grounded in three empirically driven design principles: balancing physical invariance with system specificity, accuracy with feasibility, and reliability under critical operating conditions. LUMINA integrates physics-informed neural networks, a topology-transferable architecture, and constraint-aware training objectives to enable end-to-end learning across diverse power grid datasets. Experimental results demonstrate that LUMINA efficiently solves the alternating current optimal power flow (ACOPF) problem across various grid topologies, consistently satisfying physical laws and operational constraints while maintaining high solution accuracy and robust generalization capabilities.

Technology Category

• Constraint Satisfaction and Optimization: Constraint Learning and Acquisition
• Search and Optimization: Learning to Search
• Machine Learning: Optimization

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Foundation models in general promise to accelerate scientific computation by learning reusable representations across problem instances, yet constrained scientific systems, where predictions must satisfy physical laws and safety limits, pose unique challenges that stress conventional training paradigms. We derive design principles for constrained scientific foundation models through systematic investigation of AC optimal power flow (ACOPF), a representative optimization problem in power grid operations where power balance equations and operational constraints are non-negotiable. Through controlled experiments spanning architectures, training objectives, and system diversity, we extract three empirically grounded principles governing scientific foundation model design. These principles characterize three design trade-offs: learning physics-invariant representations while respecting system-specific constraints, optimizing accuracy while ensuring constraint satisfaction, and ensuring reliability in high-impact operating regimes. We present the LUMINA framework, including data processing and training pipelines to support reproducible research on physics-informed, feasibility-aware foundation models across scientific applications.

Problem

• foundation models
• constrained scientific systems
• AC optimal power flow
• physics-informed learning
• constraint satisfaction

Innovation

• foundation models
• ACOPF
• physics-informed learning
• constraint satisfaction
• topology transfer