LAPSO: A Unified Optimization View for Learning-Augmented Power System Operations

TL;DR

LAPSO addresses the integration gap between machine learning and model-based power system operations under rising renewable uncertainty by proposing a unified, optimization-centric framework that jointly designs learning components and optimization tasks at the operation stage. It formalizes P_basic and learning-augmented P_lapso problems, introduces stability-constrained optimization (SCO) and objective-based forecasting (OBF) as prototypical LAPSO instances, and develops uncertainty-aware extensions including robust wait-and-see formulations. The authors provide open-source Python packages (lapso and pso) to automate integration of learnable components into PSO models and demonstrate the framework on an IEEE 14-bus test system with case studies showing improvements in both stability margins and economic efficiency. The work highlights end-to-end tracing of uncertainties and offers practical tools for designing task-aware, learnable constraints and forecasters that harmonize with downstream optimization. This framework enables more reliable, cost-effective, and scalable operation of future power systems with high renewable penetration.

Abstract

With the high penetration of renewables, traditional model-based power system operation is challenged to deliver economic, stable, and robust decisions. Machine learning has emerged as a powerful modeling tool for capturing complex dynamics to address these challenges. However, its separate design often lacks systematic integration with existing methods. To fill the gap, this paper proposes a holistic framework of Learning-Augmented Power System Operations (LAPSO, pronounced as Lap-So). Adopting a native optimization perspective, LAPSO is centered on the operation stage and aims to break the boundary between temporally siloed power system tasks, such as forecast, operation and control, while unifying the objectives of machine learning and model-based optimizations at both training and inference stages. Systematic analysis and simulations demonstrate the effectiveness of applying LAPSO in designing new integrated algorithms, such as stability-constrained optimization (SCO) and objective-based forecasting (OBF), while enabling end-to-end tracing of different sources of uncertainties. In addition, a dedicated Python package-lapso is introduced to automatically augment existing power system optimization models with learnable components. All code and data are available at https://github.com/xuwkk/lapso_exp.

Figures (9)

• Figure 1: Transformations of power system decision-making framework, driven by the new challenges in the grid and the new requirements of emerging ML techniques. (a). Traditional siloed model-based power system decision-making; (b). Machine-learning and optimization nexus; (c). Learning-augmented power system operation (this paper).
• Figure 2: Learning-augmented power system operation using SCO and OBF examples.
• Figure 3: The design triangle for LAPSO. Three entities, including targets, ML techniques, and PSO $\mathcal{P}_{basic}$ interact with each other.
• Figure 4: Source and propagation of uncertainties in $\mathcal{P}_{inf}^{obf/basic}$. Similar structure can be drawn for general $\mathcal{P}_{lapso}$. Stochastic formulation is taken while others, such as (distributional) robust formulations, can also be generalized to.
• Figure 5: Illustrative areas for stable and unstable operation points. The four areas can be applied to both $\mathcal{P}_{basic}$ and $\mathcal{P}_{inf}^{sco}$.

Theorems & Definitions (1)

• Definition 1