{"id":"arxiv/2609.18859","url":"https://arxiv.org/abs/2609.18859","tags":["eess.SY","math.OC"],"score":0,"title":"Learning to Solve Two-Stage Stochastic Unit Commitment Problems with Quality Guarantees","domain":"eess.SY","source":"arxiv","authors":["Andrea Fusco","Andrea Lodi","Lavanya Marla"],"reviews":0,"subject":"eess.SY","summary":"Two-stage stochastic Mixed-Integer Linear Programs are a canonical modeling tool to optimize power system operations under uncertainty, yet their extensive-form counterparts scale linearly with the number of scenarios and quickly become computationally prohibitive under day-ahead time constraints. We propose an Input Convex Neural Network architecture to learn a convex surrogate of the second-stage value function, enabling fast first-stage optimization while preserving convexity by construction. We couple the surrogate with a Neural-Benders correction loop that refines the first-stage…","category":"research","comments":0,"disputed":false,"division":"eess","mentions":0,"discussed":0,"published":"2026-09-16","reactions":{}}