Research on the Generation Model of Temperature Control Design for Surrounding Concrete of Spiral Case Based on CC-KG-RAG

Temperature control of concrete surrounding spiral case during hydropower station construction plays a critical role in ensuring structural safety and durability. However, current decision-making and management processes still suffer from fragmented knowledge sources, strong reliance on engineering experience, low efficiency in initial scheme generation, and high costs of manual verification and iterative optimization. To address these issues, this paper proposes a constraint-controlled knowledge graph–enhanced retrieval-augmented generation method (CC-KG-RAG) for intelligent temperature control scheme design. The method incorporates a dual-indexing framework combining semantic-vector retrieval with knowledge-graph retrieval, and dynamically balances semantic relevance and structured engineering knowledge through an adaptive weighted-fusion strategy. At the output stage, constraint control is applied to refine and correct generated answers. Ablation experiments confirm the contribution of each module to the overall framework, resulting in improved accuracy, stability and interpretability, while effectively suppressing hallucinations commonly observed in professional question-answering tasks of conventional large language models. Comparative experiments show that CC-KG-RAG achieves 87.3% accuracy, 84.6% recall and 92.3% credibility, demonstrating stable and well-balanced performance across the core indicators. The proposed model was applied to the spiral case surrounding concrete structure of a hydropower station to further verify its practical applicability. The results indicate that the initial temperature control scheme generated by CC-KG-RAG effectively regulates the temperature gradient. The peak temperature decreases to 43.82 °C, representing a reduction of 9.14 °C compared with the scheme without water-pipe cooling, thereby satisfying engineering requirements and temperature control standards. CC-KG-RAG reduces the number of verification iterations and the associated human resource costs during the temperature control design process. These results demonstrate its capability to support intelligent management of concrete temperature control in hydropower stations. The proposed framework therefore provides meaningful engineering value for promoting the digital and intelligent transformation of water conservancy projects.