首页|期刊导航|郑州大学学报(理学版)|基于多图嵌入式双通道图卷积的区块调整措施推荐模型

基于多图嵌入式双通道图卷积的区块调整措施推荐模型OA

Block Adjustment Measure Recommendation Model Based on Multi-graph Embedded Dual-channel Graph Convolution

中文摘要英文摘要

油田区块调整措施的智能推荐是智能油田建设的重要组成部分.针对现有方法在面临高度稀疏的油田数据时难以捕捉其中的高阶非线性关系的问题,提出一种基于多图嵌入式双通道图卷积的推荐模型.通过知识图谱嵌入方式学习区块与其属性间的潜在关系,以增强区块节点的表示能力;利用双通道图卷积网络分别捕捉异质交互图和协同图中区块与措施的高阶嵌入表示和协同嵌入表示,将两部分表示进行信息融合,实现最终的措施推荐.实验结果表明,在真实的油田开发效果评价数据集上,所提出的模型在推荐精准率、召回率和归一化折损累计增益指标上均显著优于主流基线模型.

The intelligent recommendation of oilfield block adjustment measures is an important part of the construction of intelligent oilfield.The existing methods was difficult to capture high-order nonlinear relationships in highly sparse oil field data,so a recommendation model based on multi-graph embedding dual-channel graph convolution was proposed.The latent relationship between blocks and their attributes was learnt through knowledge graph embedded to enhance the representation ability of block nodes.A du-al-channel graph convolutional network was used to capture the high-order embedded representation and collaborative embedded representation of blocks and measures in heterogeneous interaction graphs and collaborative graphs respectively,and the two representations information were fused to achieve the final measure recommendation.The experimental results showed that the proposed model was significantly bet-ter than the mainstream baseline model in the recommended precision,recall and normalized discounted cumulative gain with the real oilfield development effect evaluation data set.

王梅;尹传龙;孙洪国;朱晓丽

东北石油大学 计算机与信息技术学院 黑龙江 大庆 163318东北石油大学 计算机与信息技术学院 黑龙江 大庆 163318大庆油田勘探开发研究院开发规划研究室 黑龙江 大庆 163712东北石油大学 计算机与信息技术学院 黑龙江 大庆 163318

能源科技

智能油田双通道图卷积多图嵌入知识图谱措施推荐

intelligent oil fielddual-channel graph convolutionmulti-graph embeddingknowledge graphmeasure recommendation

《郑州大学学报(理学版)》 2026 (6)

17-24,8

国家自然科学基金项目(51774090,62076234)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)

10.13705/j.issn.1671-6841.2025066

评论