With the advancement of energy transition, large-scale energy storage stations have become crucial support for power systems, but their safety issues have become increasingly prominent. Energy storage stations feature diverse equipment types, narrow complex paths, multiple monitoring blind spots, and strong electromagnetic interference environments, making traditional safety operation and maintenance methods inadequate for rapid detection and handling of safety hazards. This paper proposes an embodied intelligence-based solution for safety operation and maintenance of energy storage stations, constructing a “fixed-mobile-aerial” multi-source universal multimodal information acquisition system, developing autonomous embodied intelligent agents adapted to the energy storage station environment, researching heterogeneous intelligent group coordination technology, and establishing an intelligent operation and maintenance computing model based on large language models, forming a complete safety operation and maintenance technology system. Through multi-agent collaborative perception and decision-making, this solution achieves comprehensive, efficient, and intelligent safety operation and maintenance of energy storage stations, providing a new technical approach for ensuring the safe operation of energy storage stations and promoting high-quality development of the energy industry.