Artificial intelligence for forecasting, optimization, and energy storage in energy communities: a review

Abstract

Energy communities (ECs) are emerging as a key paradigm for decentralized, consumer-centric energy systems, yet their operation is challenged by uncertainty, asset heterogeneity, and the need for coordinated decision-making. Artificial intelligence (AI) has been widely applied to forecasting, optimization, and management of energy storage systems (ESSs) in ECs, but existing studies remain fragmented across decision layers. This paper presents a comprehensive and systematic review of 130 peer-reviewed studies on AI-based forecasting, optimization, and ESS management in ECs. By jointly analyzing these three pillars, the review highlights how predictive, prescriptive, and adaptive components interact, and where current approaches fall short in supporting end-to-end, deployable decision pipelines. The analysis reveals that forecasting acts as a foundational layer shaping downstream decisions, while optimization and ESS control increasingly rely on learning-based and hybrid AI frameworks to manage uncertainty and multi-objective trade-offs. However, tight coupling between forecasting, optimization, and physical execution remains limited, with implications for robustness, scalability, fairness, and real-world deployment. Based on these findings, the paper synthesizes key integration challenges and outlines future research directions toward decision-centric forecasting, asset-aware control, and market-compatible AI. The review provides a unified perspective to support the design of intelligent, robust, and socio-technical EC management frameworks.