The growing demand for energy-efficient and resilient infrastructure, particularly in large-scale facilities such as data centers, necessitates innovative control strategies. This study proposes a distributed control architecture based on multi-agent systems (MAS) for air-cooled chiller systems. The framework integrates localized decision-making with coordinated optimization to enhance system-wide efficiency, scalability, and security. Unlike traditional centralized approaches, the proposed architecture employs autonomous agents for real-time monitoring, adaptation, and fault tolerance, while also meeting stringent security requirements. Experimental results demonstrate improvements of 5 to 20% in energy efficiency, 30 to 40% faster anomaly detection, and up to 30% enhancement in predictive maintenance scheduling. Compared to centralized and cloud-based methods, the proposed MAS framework offers superior adaptability, fault resilience, and cybersecurity. Future research directions include integrating hybrid AI models, generative pipelines for adaptive learning, and interoperability with emerging smart grid technologies.