Learning-to-solve unit commitment based on few-shot physics-guided spatial-temporal graph convolution network

TL;DR

This work tackles fast unit commitment by learning from a small set of optimizer-provided solutions using a physics-guided, spatio-temporal graph convolutional network (FPG-STGCN). It combines a STGCN-based UC parameterization with few-shot learning, augmented Lagrangian regularization, and a straight-through estimator to handle discrete on/off decisions, enabling feasible near-optimal solutions in one shot. A case study on a modified IEEE 30-bus system demonstrates substantially faster online decisions and improved feasibility versus traditional solvers and baselines, while ablation highlights the necessity of both the few-shot and physics-guided components. The authors note future work on theoretical sample-size guarantees for the few-shot and PGL components.

Abstract

This letter proposes a few-shot physics-guided spatial temporal graph convolutional network (FPG-STGCN) to fast solve unit commitment (UC). Firstly, STGCN is tailored to parameterize UC. Then, few-shot physics-guided learning scheme is proposed. It exploits few typical UC solutions yielded via commercial optimizer to escape from local minimum, and leverages the augmented Lagrangian method for constraint satisfaction. To further enable both feasibility and continuous relaxation for integers in learning process, straight-through estimator for Tanh-Sign composition is proposed to fully differentiate the mixed integer solution space. Case study on the IEEE benchmark justifies that, our method bests mainstream learning ways on UC feasibility, and surpasses traditional solver on efficiency.

Figures (1)

• Figure 1: Ablation test on constraint violation evolving with training epochs