Efficient scheduling of maintenance resources for hydropower units is crucial for improving maintenance efficiency and reducing costs. However, traditional methods often rely on manual experience and static scheduling rules, lacking dynamic optimization mechanisms. To address these limitations, this paper proposes an LLM-agent and knowledge graph-integrated framework for collaborative optimization of hydropower maintenance resources scheduling. First, an LLM-assisted approach is employed to construct a maintenance knowledge graph. Then, a query-aware subgraph retrieval method is designed to extract relevant triplets by identifying core entities in user queries and leveraging semantic alignment strategies for multi-hop subgraph retrieval. Subsequently, a three-stage resource allocation generation mechanism based on prompt engineering is proposed. This mechanism utilizes the retrieved knowledge subgraphs to guide the generation of diverse and structured maintenance plans. Finally, an LLM-Agent-based optimization module, incorporating a genetic algorithm, is developed to dynamically manage resource constraints and reduce costs across multiple planning alternatives. Experimental results demonstrate that the proposed framework significantly enhances the efficiency of maintenance resource allocation, offering a novel paradigm for advancing the intelligent hydropower equipment maintenance.

Hydropower Maintenance Resource Scheduling,Large Language Model,Knowledge Graph,Agent Optimization