Article Details

Abstract

To address the challenges of knowledge hallucination, poor information timeliness and low localized deployment efficiency faced by Large Language Models (LLMs) in their applications to water conservancy and other knowledge-intensive industries, this paper proposes a Knowledge-Enhanced Retrieval-Augmented Generation (KE-RAG) system architecture that integrates vector retrieval with Knowledge Graph (KG) constraints. This architecture takes the domestically developed open-source model Qwen3-32B-AWQ as the generation core, constructs a multimodal water conservancy knowledge base containing laws and regulations, technical standards, engineering cases and expert experience, and achieves high-performance localized inference through the vectorized Large Language Model (vLLM) framework. Experimental results show that on the evaluation set, this architecture reaches a professional knowledge question and answer accuracy of 91.5%, an average end-to-end response latency of less than 500 ms and a throughput nearly four times higher than that of standard deployment, with all metrics significantly outperforming those of baseline models, which provides a reference solution for the application of Large Language Models in the water conservancy industry.

Keywords

Large Language Model; retrieval-augmented generation; Knowledge Graph; water conservancy knowledge question and answer; high-performance inference; localized deployment

References

1. [1] Xu Yuemei, Hu Ling, Zhao Jiayi, etc The Technical Application Prospects and Risk Challenges of Large Language Models [J]. Computer Applications, 2024, 44 (6): 1655-1662
2. [2] Li Zijun, Xiao Hui, Li Xuefeng A review of retrieval enhanced generation techniques for knowledge intensive tasks [J]. Microelectronics and Computer, 2025, 42 (10): 48-65
3. [3] Feng Jun, Chang Yanghong, Lu Jiamin, etc Construction and Application of Water Engineering Scheduling Knowledge Graph Based on Large Language Model [J]. Computer Science and Exploration, 2024, 18 (6): 1637-1647
4. [4] Yang Yangrui, Dong Fangning, Wang Pengfei, etc Automated Construction of Quality Concept Model and Knowledge Graph for Water Conservancy Engineering Operation Management Based on Large Language Model [J]. Journal of Water Resources, 2025, 56 (5): 634-645
5. [5] Wang Yaozu, Li Qing, Dai Zhangjie, etc Current Status and Trends of Large Language Model Research [J]. Journal of Engineering Science, 2024, 46 (8): 1411-1425
6. [6] Vincent, Qian Li, Hu Maodi, etc A review of research progress on question answering technology based on big language models [J]. Data Analysis and Knowledge Discovery, 2024, 8 (6): 16-29
7. [7] Zhu Qiaoli, Wu Shangyou, Cheng Jie, etc Research on the Quality Evaluation Method of Large Model Water Conservancy Knowledge Q&A Based on RAG [C]//2024 China Water Conservancy Academic Conference Xi'an: Chinese Water Conservancy Society, 2024: 473-478
8. [8] Jiang Xinlan, Yu Liyuan, Lin Lin, etc Research on the Application of Intelligent Q&A System in Querying Regulations and Rules in the Water Conservancy Field [C]//2024 China Water Conservancy Academic Conference Xi'an: Chinese Water Conservancy Society, 2024: 526-537
9. [9] Translated by Zhang He, Wang Xin, Han Lifan, etc Research on Question Answering System Based on Large Language Model Fusion Knowledge Graph [J]. Computer Science and Exploration, 2023, 17 (10): 2377-2388
10. [10] Ji Weiyu, Zhang Yong, Jiang Wei Research on the Technology of Knowledge Enhancement Large Model Generation Framework for Business Requirements [J]. Software, 2024, 45 (5): 158-160
11. [11] KWON W，LI Z H，ZHUANG S Y，et al. Efficient Memory Management for Large Language Model Serving with PagedAttention [C]//SOSP' 23: Proceedings of the 29th Symposium on Operating Systems Principles. New York：ACM，2023：611-626.

Funding

Hunan Water Conservancy Science and Technology Project (XSKJ2024064-30)

About the authors

Ding Yuliang (1995-), male, Han nationality, from Yiyang, Hunan, intermediate engineer, bachelor's degree, research direction: application of large-scale models and software development in the water conservancy industry; Corresponding author: Peng Xiulian (1988-), female, Han nationality, from Loudi, Hunan, senior engineer, undergraduate, research direction: water conservancy management.