摘要：Liquid loading in gas wells remains one of the most critical challenges in natural gas production, and plunger lift is widely recognized as one of the most effective technologies for gas well deliquification. However, existing fault diagnosis methods for plunger lift systems largely depend on manual inspection, which is labor-intensive, timeconsuming, and lacks automation and intelligence. Unsupervised or semi-supervised approaches often suffer from poor robustness. To foster the deep integration of information technology with gas production processes and support the intelligent and digital transformation of gas well operations, this study proposes a novel method that incorporates domain knowledge into the diagnostic model. A domain-specific knowledge graph for plunger lift systems is constructed to formally represent expert knowledge, including concepts, entities, and their interrelations. By integrating this knowledge graph into a conventional deep neural network using graph embedding techniques, we develop a hybrid neural network capable of performing intelligent fault diagnosis driven by both physical mechanisms and data patterns. This integration significantly enhances the model＇s ability to understand and leverage domain expertise while improving the interpretability of its predictions. Experimental results demonstrate that the proposed model achieves high diagnostic performance across various working conditions, with an average accuracy of 0.903 and peak values of 0.966, 0.958, and 0.96 for Precision, Recall, and AUC, respectively. Field validation confirms that the model can accurately identify actual operational states and provide targeted optimization recommendations for different fault scenarios.

关键字：Plunger lift; Gas well deliquification; Domain-specific knowledge graph; Hybrid neural network; Real-time fault diagnosis

ISSN号：2352-4847

卷、期、页：卷: 14页: 5137-5156

发表日期：2025-12-01

期刊分区（SCI为中科院分区）：三区

收录情况：SCI（科学引文索引印刷版）,EI（工程索引）,SCIE（科学引文索引网络版）

发表期刊名称：ENERGY REPORTS

参与作者：龚航飞

通讯作者：邢志晟,路鑫

第一作者：梁星原,王广祖,韩国庆

论文类型：期刊论文

论文概要：梁星原,邢志晟,龚航飞,王广祖,路鑫,韩国庆，Intelligent fault diagnosis method for plunger lift systems integrating knowledge graph constraints，ENERGY REPORTS，2025，卷: 14页: 5137-5156

论文题目：Intelligent fault diagnosis method for plunger lift systems integrating knowledge graph constraints