<p>Cyber-physical energy systems (CPES) are the core components of an energy system. A CPES is characterized by a high level of complexity and uncertainty, while ensuring a more resilient and efficient distributed energy system, including robust energy security and economic viability. This work presents a novel and effective approach for anomaly detection in CPES systems using advanced tree-based machine learning models. The proposed approach is centralized with a minimal set of features that achieve a maximum detection rate and system performance. The results emphasize the effectiveness of ML methods in this domain, showcasing their ability to handle complex, high-dimensional datasets while providing interpretability and efficiency. The study also highlights associated challenges such as scalability and adversarial resilience, proposing future research directions such as hybrid models and interpretable AI for real-world deployment.</p>