To overcome limitations of existing methods, which typically struggle to enforce feasibility constraints or leverage the knowledge available in the mathematical model, we propose a learning framework based on identifying the relevant sets of active constraints. Using the active sets as an intermediate step enables efficient recovery of the optimal solution, inherently accounts for relevant safety constraints and provides more interpretable results. Further, while the number of possible active sets is combinatorial in the system size, the number of practically relevant active sets can be small, which make them simpler objects to learn. To identify the relevant active sets, we propose a streaming algorithm with rigorous probabilistic performance guarantees. The algorithm is demonstrated using the optimal power flow (OPF) with renewable energy production as an example. We establish that the number of active sets is typically small in this problem, and discuss practical interpretations for power system operation.
Discovery Building, Orchard View Room