基于融合评价指标 BERT-RGCN 的油田评价区块调整措施推荐方法

Recommendation method for adjustment measures of oilfield evaluation blocks based on the fusion evaluation index BERT-RGCN

摘要

为解决油田领域区块调整措施推荐过程中存在的样本数据稀疏和语义特征复杂等问题，提出基于融合评价指标（EI）的变换器双向编码（BERT）与关系图卷积神经网络（RGCN）的油田评价区块调整措施推荐方法（EI-BERTRGCN 方法）。根据评价指标、评价区块及措施之间的交互信息构建异构图，利用BERT 模型生成评价指标、评价区块及措施术语词向量，共同作为输人词向量，将融合评价指标信息的异构图和输入词向量放入 RGCN 模型训练，学习评价区块的有效表征；在某油田评价区块提供的数据集上进行实验对比。 结果表明：EI-BERT-RGCN 方法能够捕捉文本中隐含的复杂语义并缓解数据稀疏问题，能更好理解未观察到的评价指标与调整措施之间的潜在关系，提升节点的表示质量。EI-BERT-RGCN 模型在精确率、召回率、 F₁ 分数及 ROC 曲线下面积等评价指标上优于其他基准模型，在保持较高精确率的同时，展现更好的泛化能力和鲁棒性。该结果为油田评价区块调整措施推荐提供参考。

Abstract

In order to solve the problems of sparse sample data and complex semantic features in the process of recommending block adjustment measures in the oilfield field, an adjustment measure recommendation method based on the Evaluation Index-empowered Bidirectional Encoder Representations from Transformers(BERT) and Relational Graph Convolutional Networks (RGCN) (EI-BERT-RGCN method) is proposed. A heterogeneous map is constructed based on the interaction information among evaluation indicators, evaluation blocks and measures, and the BERT model is used to generate evaluation indicator, evaluation block and measure term word vectors, which are jointly used as input word vectors, and the heterogeneous map and input word vectors fusing the information of evaluation indicators are put into the RGCN model for training to learn an effective characterization of the evaluation blocks; experimental comparisons are conducted on the dataset provided by the evaluation blocks of an oil field. The results show that: the EI-BERT-RGCN method can capture the complex semantics implied in the text and alleviate the data sparsity problem, better understand the potential relationship between unobserved evaluation indicators and adjustment measures, and improve the representation quality of the nodes. The EI-BERT-RGCN model outperforms the other benchmark models in terms of evaluation metrics such as the precision rate, the recall rate, the F₁ score and the area under the ROC curve the EI-BERT-RGCN model demonstrates better generalization ability and robustness while maintaining a high precision rate. The results provide a reference for recommending adjustment measures for oilfield evaluation blocks.

关键词

异构图 / 变换器双向编码（BERT） / 预训练模型 / 关系图卷积神经网络（RGCN） / 推荐算法 / 措施推荐 / 油田评价区块

Key words

heterogeneous graph / Bidirectional Encoder Representations from Transformers (BERT) / pre-trained models / Relational Graph Convolutional Networks (RGCN) / recommendation algorithms / measure recommendation / oil field evaluation blocks

参考文献

[1] 周志军, 张国青, 崔春雪, 等. 页岩储层孔隙结构表征及物性下限确定方法及应用[J]. 特种油气藏, 2024, 31(4):96-102.

[2] ZHOU Zhijun, ZHANG Guoqing, CUI Chunxue, et al. Methods and applications for characterizing pore structure and determining physical property lower limit in shale reservoirs[J]. Special Oil & Gas Reservoirs, 2024, 31(4):96-102.

[3] 王双龙, 岳宝林, 祝晓林, 等. 渤海锦州南变质岩潜山油田储层分类及其剩余油分布规律[J]. 东北石油大学学报, 2024, 48(4):89-99.

[4] WANG Shuanglong, YUE Baolin, ZHU Xiaolin, et al. Reservoir classification and remaining oil distribution patterns of Jinzhou south metamorphic buried mountain oilfield in Bohai Sea[J]. Journal of Northeast Petroleum University, 2024, 48(4):89-99.

[5] 王洪亮, 李宁, 李欣, 等. 基于数据驱动的致密油藏产能预测方法[J]. 特种油气藏, 2025, 32(2):82-88.

[6] WANG Hongliang, LI Ning, LI Xin, et al. Productivity prediction method of tight reservoir based on data-driven approach[J]. Special Oil & Gas Reservoirs, 2025, 32(2):82-88.

[7] 张凤远, 邹林君, 崔维, 等. 基于压裂液返排数据的页岩油气藏裂缝参数反演方法[J]. 东北石油大学学报, 2022, 46(1):76-87.

[8] ZHANG Fengyuan, ZOU Linjun, CUI Wei, et al. Inverse method based on fracturing fluid flowback data for estimating fracture properties in shale reservoirs[J]. Journal of Northeast Petroleum University, 2022, 46(1):76-87.

[9] 彭神奇. 大牛地气田补孔措施效益后评价及效益界限产量研究[J]. 中外能源, 2020, 25(6):53-56.

[10] PENG Shenqi. Benefit evaluation and benefit limit production research of reperforating measures in Daniudi Gas Field[J]. Sino-Global Energy, 2020, 25(6):53-56.

[11] 苑登御, 侯吉瑞, 宋兆杰, 等. 塔河油田缝洞型碳酸盐岩油藏注水方式优选及注气提高采收率实验[J]. 东北石油大学学报, 2015, 39 (6):102-110.

[12] YUAN Dengyu, HOU Jirui, SONG Zhaojie, et al. Optimization of water injection methods and N2 flooding for EOR in Tahe fractured-vuggy carbonate reservoir[J]. Journal of Northeast Petroleum University, 2015, 39(6):102-110.

[13] 欧阳雨薇, 龙明, 张吉磊, 等. 渤海河流相 Q 油田提液挖潜规律分析[J]. 石油化工高等学校学报, 2023, 36(1):32-39.

[14] OUYANG Yuwei, LONG Ming, ZHANG Jilei, et al. Law of liquid production in increase in fluvial facies reservoir of Bohai Q Oilfield[J]. Journal of Petrochemical Universities, 2023, 36(1):32-39.

[15] 刘丹阳. A 油田压裂井效果分析与研究[J]. 石化技术, 2015, 22(8):174.

[16] LIU Danyang. Fraturing wells in A Oilfield[J]. Petrochemical Industry Technology, 2015, 22(8):174.

[17] 黄易泽, 刘国栋, 李洲, 等. 安塞油田致密砂岩油藏补孔压裂措施效果分析[J]. 石油地质与工程, 2024, 38(2):88-92.

[18] HUANG Yize, LIU Guodong, LI Zhou, et al. Analysis of perforation fracturing measures effect in tight sandstone reservoir of Ansai Oilfield[J]. Petroleum Geology and Engineering, 2024, 38(2):88-92.

[19] 高雅田, 李春生, 富宇. 基于知识解析的油田动态开发辅助决策系统研究[J]. 科学技术与工程, 2010, 10(11):2767-2771.

[20] GAO Yatian, LI Chunsheng, FU Yu. Research on assistant decision system for oilfield dynamic development based on analysis of knowledge[J]. Science Technology and Engineering, 2010, 10(11):2767-2771.

[21] 孟军晓. 先进技术在石油开采率提升中的应用分析[J]. 中国石油和化工标准与质量, 2025, 45(17):146-148.

[22] MENG Junxiao. Application analysis of advanced technologies in improving oil recovery rate[J]. China Petroleum and Chemical Standard and Quality, 2025, 45(17):146-148.

[23] MA S, ZHANG C, LIU X. A review of citation recommendation:from textual content to enriched context[J]. Scientometrics, 2020, 122(3):1445-1472.

[24] LU Y, HE Y, CAI Y, et al. Time-aware neural collaborative filtering with multi-dimensional features on academic paper recommendation[C]//International Conference on Computer Supported Cooperative Work in Design.Dalian, China:IEEE,2021:1052-1057.

[25] JEONG C, JANG S, PARK E, et al. A context-aware citation recommendation model with BERT and graph convolutional networks[J]. Scientometrics, 2020, 124(1):1907-1922.

[26] COVINGTON P, ADAMS J, SARGIN E. Deep neural networks for YouTube recommendations[C]//ACM Conference on Recommender Systems.Boston,MA, USA:ACM,2016:191-198.

[27] CHENG H T, KOC L, HARMSEN J, et al. Wide & deep learning for recommender systems[C]//Proceedings of the 1st Workshop on Deep Learning for Recommender Systems.Boston,MA, USA:ACM,2016:7-10.

[28] RUMELHART D E, HINTON G E, WILLIAMS R J. Learning representations by back-propagating errors[J]. Nature, 1986, 323 (6088):533-536.

[29] LYU Z, WU Y, LAI J, et al. Knowledge enhanced graph neural networks for explainable recommendation[J]. IEEE Transactions on Knowledge and Data Engineering, 2022, 35(5):4954-4968.

[30] SARZYNSKA-WAWER J, WAWER A, PAWLAK A, et al. Detecting formal thought disorder by deep contextualized word representations[J]. Psychiatry Research, 2021,304:114135.

[31] RADFORD A, WU J, CHILD R, et al. Language models are unsupervised multitask learners[J]. Open AI Blog, 2019, 1(8):9.

[32] DEVLIN J, CHANG M W, LEE K, et al. BERT:pre-training of deep bidirectional transformers for language understanding[C]// Proceedings of the 2019 Conference of the North American Chapter of the Association for Computational Linguistics:Human Language Technologies. Minneapolis, Minnesota:ACM,2019:4171-4186.

[33] 郭军辉, 郑宪宝, 王治国, 等. 大庆长垣油田水驱开发技术智能化实践与展望[J]. 大庆石油地质与开发, 2024, 43(3):203-213.

[34] GUO Junhui, ZHENG Xianbao, WANG Zhiguo, et al. Intelligent practice and prospects of water flooding development technology in Daqing Placanticline Oilfield[J]. Petroleum Geology & Oilfield Development in Daqing, 2024, 43(3):203-213.

[35] 王晓丽, 邓达康, 孟祥龙, 等. 基于领域本体的油气勘探开发知识获取模式及实现[J]. 东北石油大学学报, 2016, 40(4):74-79.

[36] WANG Xiaoli, DENG Dakang, MENG Xianglong, et al. Oil and gas exploration and production knowledge processing mode and implementation based on domain ontology[J]. Journal of Northeast Petroleum University, 2016, 40(4):74-79.

[37] 高丽洁. 基于模式匹配与最近邻算法的油井维护措施推荐[D]. 北京: 中国石油大学(北京),2018:28-34.

[38] GAO Lijie. Measures recommendation of oil well maintenance based on pattern matching and nearest neighbor algorithm[D]. Beijing: China University of Petroleum(Beijing),2018:28-34.

[39] 李权, 李金蔓, 张艳英, 等. 知识图谱在油井措施推荐方面的应用[J]. 石化技术, 2022, 29(3):124-125.

[40] LI Quan, LI Jinman, ZHANG Yanying, et al. Application of knowledge atlas in oil well measure recommendation[J]. Petrochemical Industry Technology, 2022, 29(3):124-125.

[41] SCHLICHTKRULL M, KIPF T N, BLOEM P, et al. Modeling relational data with graph convolutional networks[C]//European Semantic Web Conference. Cham: Springer International Publishing,2018:593-607.

[42] 成全, 佘德昕. 融合患者体征与用药数据的图神经网络药物推荐方法研究[J]. 数据分析与知识发现, 2022, 6(9):113-124.

[43] CHENG Quan, SHE Dexin. Drug recommendation based on graph neural network with patient signs and medication data[J]. Data Analysis and Knowledge Discovery, 2022, 6(9):113-124.

[44] HUANG C, LI Y, LOY C C, et al. Learning deep representation for imbalanced classification[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE,2016:5375-5384.

[45] LU J, CHEN Z, DENG X L. A graph convolutional neural network model based on fused multi-subgraph as input and fused feature information as output[J]. Engineering Applications of Artificial Intelligence, 2025,139:109542.

[46] BAO D, QIN D, HONG L, et al. Multi-channel text classification model based on ERNIE[C]//Proceedings of the 2022 11th International Conference on Computing and Pattern Recognition. Beijing,China: ACM,2022:321-327.

[47] WANG M, XU J. Research on Chinese short text classification based on ERNIE-TEXTCNN model[C]//Third International Conference on Advanced Algorithms and Neural Networks(AANN 2023). Qingdao,China:SPIE, 2023,12791:315-319.

[48] MACEDO J B, MOURA M D C, AICHELE D, et al. Identification of risk features using text mining and BERT-based models:application to an oil refinery[J]. Process Safety and Environmental Protection, 2022,158:382-399.

[49] LIU Q, XIAO K, QIAN Z. A hybrid re-fusion model for text classification[J]. Scientific Reports, 2025, 15(1):9333.

[50] CUI Y, CHE W, LIU T, et al. Pre-training with whole word masking for Chinese BERT[J]. IEEE/ACM Transactions on Audio, Speech,and Language Processing, 2019,29:3504-3514.

基金资助

国家自然科学基金项目(52274037)

黑龙江省科技创新基地项目(JD24A009)

黑龙江省自然科学基金项目(LH2024F005)

黑龙江省博士后科研启动金资助项目(LBH-Q20080)