Linear and Second-order-cone Valid Inequalities for Problems with Storage

TL;DR

This paper develops new linear and second-order cone inequalities based on the disjunctive structure of battery operation, improving the tractability of models that prevent simultaneous charging and discharging in energy systems.

Contribution

It introduces a systematic method to derive valid inequalities that define the convex hull of feasible battery operations, including a family that generalizes existing formulations.

Findings

Facets of the convex hull are characterized by new linear inequalities.

Second-order cone inequalities effectively reduce simultaneous charge and discharge.

Structured relaxations improve model tractability and accuracy in energy system optimization.

Abstract

Batteries are playing an increasingly central role as distributed energy resources in the shift toward power systems dominated by renewable energy sources. However, existing battery models must invariably rely on complementarity constraints to prevent simultaneous charging and discharging, rendering models of a disjunctive nature and NP-hard. In this paper, we analyze the disjunctive structure of the battery's feasible operational set and uncover a submodularity property in its extreme power trajectories. Leveraging this structure, we propose a systematic approach to derive linear valid inequalities that define facets of the convex hull of the battery's feasible operational set, including a distinguished family that generalizes and dominates existing formulations in the literature. To evaluate the practical utility of these inequalities, we conduct computational experiments on two representative problems whose continuous relaxations frequently result in simultaneous charge and discharge: energy arbitrage under negative prices and set-point tracking. For the latter, we further introduce second-order cone inequalities that remarkably and efficiently reduce simultaneous charging and discharging. Our results highlight the potential of structured relaxations to enhance the tractability and fidelity of battery models in optimization-based energy system tools.