Low fidelity models improved using self supervised tuning with high fidelity feedback
Self-Supervised Tuning of Low-Fidelity Optimization Models via High-Fidelity Feedback
Machine Learning
Summary
Low-fidelity models are fast but often inaccurate versions of complex systems. The authors propose a method that tunes these fast models by using feedback from more accurate high-fidelity models, without needing costly exact solutions. They do this by embedding the low-fidelity model in a system that learns from the high-fidelity feedback to improve its predictions. This approach works well on problems like power flow and gas networks, and generalizes to new conditions and large systems.
What this means in practice
- •For power grid operators: Improve fast power flow models to better match high-quality operational constraints and increase solution feasibility across conditions.
- •For gas network managers: Enhance gas flow optimization models efficiently using high-fidelity feedback without expensive exact computations.
Authors
Gonzalo E. Constante-Flores, André H. Quisaguano-Paredes, Asha Ramanujam, Rishabh Gupta, Can Li
Abstract
Low-fidelity (LF) optimization models are widely used in large-scale operations because they can be solved efficiently and reliably, but they often exhibit systematic errors relative to high-fidelity (HF) formulations that better capture physical and operational constraints. Existing learning-based approaches typically rely on supervised or semi-supervised training with HF solutions, which are expensive to compute and may not align with HF operational objectives. We propose a self-supervised framework that parameterizes LF optimization models directly from HF feedback. The low-fidelity problem is embedded as a differentiable optimization layer and trained end-to-end, while a HF correction problem evaluates its solution and provides gradient information through its dual variables, without requiring HF solution labels. We evaluate the framework on optimal power flow, optimal gas flow, and pooling problems, spanning different sources of low- and high-fidelity mismatch. Across all three applications, the self-supervised approach improves objective quality and feasibility over untuned LF models and generally outperforms semi-supervised training. The learned parameterizations also generalize to unseen operating conditions and, for network applications with up to 30,000 nodes, to systems not encountered during training.