Mapping High-Performance Regions in Battery Scheduling across Data Uncertainty, Battery Design, and Planning Horizons

A team of six researchers, including Jaime de Miguel Rodriguez and Artjom Vargunin, has published a comprehensive study on arXiv that provides a new framework for optimizing battery energy storage scheduling. Using synthetic datasets to systematically explore the interplay between forecast uncertainty, battery design (specifically C-rate), data profiles, and planning horizon length, the research introduces the concept of an 'effective horizon.' This is the specific look-ahead length in Model Predictive Control (MPC) beyond which additional forecast information yields diminishing returns for operational performance.

The study's major contribution is a practical map that correlates battery types, uncertainty levels, and data characteristics to their optimal horizon lengths. For industrial operators, this means they can significantly reduce the computational burden of their scheduling algorithms—by stopping calculations at the effective horizon—without sacrificing financial performance. The paper also quantifies how forecast errors directly lead to revenue losses, a critical insight even for fast-responding batteries. Furthermore, the framework establishes a foundation for future machine learning models that can dynamically predict optimal horizons from dataset parameters, enabling continuous, low-compute optimization in real-world grid and industrial storage applications.