Methods for the Operational Management of Battery Storage Using Machine Learning in Microgrids

Systems (BESS) provide essential flexibility to address these challenges, yet their operational management remains a complex task under technical constraints and uncertainty. This paper presents a systematic performance evaluation of operational control strategies for battery energy storage systems in smart grids. The main contribution lies in the comparative assessment of conventional rule-based and optimization-based approaches alongside machine-learningbased methods, including supervised learning and reinforcement learning. All strategies are evaluated under identical and realistic boundary conditions using a representative gridconnected microgrid case study. Battery operation is formulated as a sequential decisionmaking problem, enabling a transparent comparison across methods using consistent performance metrics related to economic efficiency, grid interaction, battery stress, robustness, and computational effort. The results indicate that machine-learning-based strategies, particularly reinforcement learning, can achieve improved overall performance and robustness while significantly reducing online computational requirements. By focusing on a comprehensive and reproducible performance analysis rather than on a single control algorithm, this work provides practical insights into the selection and application of operational strategies for battery storage systems in smart grids.

.