SQL4Domain: 面向电力垂类领域的高可靠Text-to-SQL生成与校验框架

doi:10.3969/j.issn.1000-5641.2026.05.016

摘要/Abstract

摘要：

近年来, 随着大语言模型 (Large Language Model, LLM) 的快速发展, Text-to-SQL任务取得了突破性进展. 然而, 在垂类领域 (如金融、医疗、电力等行业) 中, 由于文本问句表达复杂且数据库结构专业性强, Text-to-SQL任务仍然面临重大挑战. 本文提出了一种面向垂类领域的高可靠Text-to-SQL智能解析框架SQL4Domain, 以电力领域为典型应用案例. 该框架首先利用垂类领域知识库对用户问句进行意图识别和关键实体抽取, 随后结合领域知识提示进行SQL (Structured Query Language)生成, 最后通过多策略知识验证提升SQL的准确性和可靠性. 在电力领域数据集上的实验结果表明, SQL4Domain在执行准确率上显著优于现有方法, 相较于主流的DAIL方案整体性能提升了21.65%.

关键词: Text-to-SQL, 垂类大数据, 大语言模型

Abstract:

Owing to the rapid development of large language model (LLM), Text-to-SQL tasks have progressed significantly in recent years. However, applying Text-to-SQL to specialized domains such as finance, healthcare, and power engineering remains challenging owing to complex user queries and highly specialized database schemas. This paper proposes a highly reliable Text-to-SQL parsing framework named SQL4Domain, with the power industry considered as a typical application scenario. The framework first utilizes domain-specific knowledge bases to implement user-query intent recognition and key entity extraction. Subsequently, it generates structured query language (SQL) statements combined with domain knowledge prompts. Finally, it improves the accuracy and reliability of SQL statements via multi-strategic knowledge verification. Experiments on the power-industry dataset reveal that SQL4Domain significantly outperforms existing methods in terms of execution accuracy, with a performance improvement of 21.65% compared with the mainstream DAIL method.

Key words: Text-to-SQL, domain-specific big data, large language models

中图分类号:

TP391

张远来, 汪庆宁, 高明. SQL4Domain: 面向电力垂类领域的高可靠Text-to-SQL生成与校验框架[J]. 华东师范大学学报（自然科学版）, 2026, 2026(5): 196-207.

Yuanlai ZHANG, Qingning WANG, Ming GAO. SQL4Domain: A highly reliable Text-to-SQL generation and verification framework for domain-specific electric power data[J]. J* E* C* N* U* N* S*, 2026, 2026(5): 196-207.

图/表 7

图1

表1

表2

表3

表4

图2

图3

参考文献 22

1	Ahmed M M R, Mirsaeidi S, Koondhar M A, et al.. Mitigating uncertainty problems of renewable energy resources through efficient integration of hybrid solar PV/wind systems into power networks. IEEE Access, 2024, 12, 30311- 30328.
2	Ni B Y, Cai X L, Shen Z J, et al.. Intelli-Dispatch-SQL: an LLM-based agent for reliable Text-to-SQL in power dispatching. Energy and AI, 2025, 22, 100591.
3	Liu X Y, Shen S Y, Li B Y, et al.. A survey of Text-to-SQL in the era of LLMs: where are we, and where are we going?. IEEE Transactions on Knowledge and Data Engineering, 2025, 37 (10): 5735- 5754.
4	Guo A B, Zhao X, Ma W B.. ER-SQL: learning enhanced representation for Text-to-SQL using table contents. Neurocomputing, 2021, 465, 359- 370.
5	Gao D W, Wang H B, Li Y L, et al.. Text-to-SQL empowered by large language models: a benchmark evaluation. Proceedings of the VLDB Endowment, 2024, 17 (5): 1132- 1145.
6	Bai J Z, Bai S, Chu Y F, et al. Qwen technical report [PP/OL]. V1. arXiv (2023-09-28)[2026-07-12]. https://arxiv.org/abs/2309.16609.
7	齐冬莲, 李启, 陈毅, 等.. 面向电力设备智慧运维的数字孪生关键技术研究现状与展望. 高电压技术, 2026, 52 (4): 1501- 1517.
8	夏元轶, 滕昌志, 徐波, 等.. 面向电力系统差异化业务的数据处理架构与低时延传输方法研究. 南京邮电大学学报(自然科学版), 2024, 44 (5): 37- 46.
9	李莉, 朱永利, 宋亚奇.. 电力设备监测数据的流式计算与动态可视化展示. 电力建设, 2017, 38 (5): 91- 97.
10	王德文, 李静芳.. 变电设备状态监测大数据的查询优化方法. 电力系统自动化, 2017, 41 (2): 165- 172.
11	闫玮丹, 齐冬莲, 闫云凤, 等.. 面向电力领域的知识图谱与大模型融合关键技术及其典型应用. 高电压技术, 2025, 51 (4): 1747- 1762.
12	Tang Y C, Han H Y, Yu X M, et al. An intelligent question answering system based on power knowledge graph [C]//2021 IEEE Power & Energy Society General Meeting (PESGM). IEEE, 2021: 1-5.
13	Cheng S, Cheng Q N, Jin L B, et al. SQLord: a robust enterprise Text-to-SQL solution via reverse data generation and workflow decomposition [C]//Companion Proceedings of the ACM on Web Conference 2025. ACM, 2025: 919-923.
14	Hong Z J, Yuan Z, Zhang Q G, et al.. Next-generation database interfaces: a survey of LLM-based Text-to-SQL. IEEE Transactions on Knowledge and Data Engineering, 2025, 37 (12): 7328- 7345.
15	Wang B L, Shin R, Liu X D, et al. RAT-SQL: relation-aware schema encoding and linking for Text-to-SQL parsers [C]//Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics. Association for Computational Linguistics, 2020: 7567-7578.
16	Xu X J, Liu C, Song D. SQLNet: generating structured queries from natural language without reinforcement learning [PP/OL]. V1. arXiv (2017-11-13)[2026-07-12]. https://arxiv.org/abs/1711.04436.
17	Liu G L, Tan Y Z, Zhong R C, et al. Solid-SQL: enhanced schema-linking based in-context learning for robust Text-to-SQL [C]//Proceedings of the 31st International Conference on Computational Linguistics. Association for Computational Linguistics, 2025: 9793-9803.
18	Yu T, Zhang R, Yang K, et al. Spider: a large-scale human-labeled dataset for complex and cross-domain semantic parsing and Text-to-SQL task [C]//Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing. Association for Computational Linguistics, 2018: 3911-3921.
19	Li H Y, Zhang J, Li C P, et al. RESDSQL: decoupling schema linking and skeleton parsing for Text-to-SQL [C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2023, 37(11): 13067-13075.
20	Houlsby N, Giurgiu A, Jastrzebski S, et al. Parameter-efficient transfer learning for NLP [C]//Proceedings of the 36th International Conference on Machine Learning. PMLR, 2019, 97: 2790-2799.
21	Hu E J, Shen Y L, Wallis P, et al. LoRA: low-rank adaptation of large language models [C]//The Tenth International Conference on Learning Representations. 2022: 1-26.
22	Zheng Y W, Zhang R C, Zhang J H, et al. LlamaFactory: unified efficient fine-tuning of 100 + language models [C]//Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations). Association for Computational Linguistics, 2024: 400-410.

模型	方法	EX	EM
Qwen-plus	DAIL	0.642	0.165
Qwen-plus	–	0.195	0.018
Qwen2.5-14B-Instruct	SFT	0.715	0.470
Qwen2.5-14B-Instruct	SQL4Domain(ours)	0.781	0.451

模型	方法	EX
Qwen2.5-14B-Instruct	本文方法	0.781
	w/o Correct	0.760
	w/o Check	0.766
	w/o Schema Linking	0.751

模型	纠错次数$ c $	EX
Qwen2.5-14B-Instruct	0(Baseline)	0.760
	1	0.775
	2	0.781
	3	0.781

方法	Precision	Recall	$ {F}_{1} $
Base Method	0.68	0.59	0.63
+ Consistency Check	0.80	0.59	0.68
+ Heuristic Rules	0.74	0.78	0.76

[1]	纪坤华, 金浩阳, 刘文辉, 黄定江. 大语言模型与多智能体赋能新型电力系统[J]. 华东师范大学学报（自然科学版）, 2026, 2026(5): 178-195.
[2]	宫泽正, 沈富可, 魏同权. 基于大语言模型的测试用例图谱生成研究[J]. 华东师范大学学报（自然科学版）, 2026, 2026(4): 143-153.
[3]	马毅鸣, 林欣, 倪琴, 余杨泽, 邓赐平, 贺樑. 针对大语言模型的社会认知评测框架和问答基准[J]. 华东师范大学学报（自然科学版）, 2026, 2026(4): 102-111.
[4]	谢林娜, 陆雪松. 基于学生开源社区行为的数字岗位就业预测[J]. 华东师范大学学报（自然科学版）, 2025, 2025(5): 87-98.
[5]	许立君, 杨丽, 黄子祎. 大语言模型与开源生态协同的人工智能课程创新探索与研究[J]. 华东师范大学学报（自然科学版）, 2025, 2025(5): 66-75.
[6]	庞瑞洋, 陆雪松. 基于智能体的可交互数据结构和算法可视化实现[J]. 华东师范大学学报（自然科学版）, 2025, 2025(5): 32-42.
[7]	何德鑫, 韩凡宇, 王伟. 大语言模型在开源项目主题标注中的应用与评估研究[J]. 华东师范大学学报（自然科学版）, 2025, 2025(5): 14-24.
[8]	陈郅睿, 陆雪松. 基于开源代码大语言模型提示的学生代码修复[J]. 华东师范大学学报（自然科学版）, 2024, 2024(5): 93-103.
[9]	寇思佳, 闫凤云, 马晶. 国内大语言模型在学科知识图谱自动标注上的应用——以道德与法治和数学学科为例[J]. 华东师范大学学报（自然科学版）, 2024, 2024(5): 81-92.
[10]	刘佳, 孙新, 张宇晴. 知识图谱与大语言模型协同的教育资源内容审查[J]. 华东师范大学学报（自然科学版）, 2024, 2024(5): 57-69.