基于知识图谱的测井储层推荐算法研究

来源：计算机技术与发展杂志2023年第04期北京时间：

作者：尚福华;张月霞;曹茂俊

单位：东北石油大学 计算机与信息技术学院,黑龙江 大庆 163318 Author(s): SHANG Fu-hua;ZHANG Yue-xia;CAO Mao-jun School of Computer and Information Technology,Northeast Petroleum University,Daqing 163318,China 关键词: 知识图谱;知识图谱补全;测井储层信息;注意力机制;推荐算法 Keywords: knowledge graph;completion of knowledge graph;logging reservoir parameters;attention mechanism;recommendation algorithm 分类号: TP391 DOI: 10. 3969 / j. issn. 1673-629X. 2023. 04. 020 摘要: 针对实际应用中测井解释人员由于经验不足以及测井解释模型所选取的参数不当、差错等,所造成的测井处理解释结果的评价精度不准等问题,提出了一种基于知识图谱的测井储层推荐算法。 首先,构建测井领域知识图谱对测井解释操作人员和其测井储层特性信息进行统一表示;其次,通过加入注意力机制和 TransR 算法对测井领域知识图谱进行补全,再通过连接机制把推荐算法与知识图谱结合起来;最后,根据潜在特征预测并推荐测井领域中储层参数信息。 在测井领域数据集上的实验结果表明,该推荐算法在 Top-K 推荐中 K = 15 时,其准确率、召回率和归一化折损累计增益三个指标比同类算法分别提高了 6. 21% 、8. 05% 、4. 82% ,并能够高精度预测和推荐测井领域储层参数信息以及更准确地判断储层油气状况,从而揭示出基于知识图谱构建测井储层推荐算法的研究方案是可行的。 Abstract: In order to solve the problems of inaccurate evaluation accuracy of logging interpretation results caused by inexperience oflogging interpreters and improper parameters and errors of logging interpretation model in practical application, a reservoirrecommendation algorithm based on knowledge graph is proposed. Firstly,the logging domain knowledge map is constructed to representthe logging interpretation operators and their logging reservoir characteristics uniformly. Secondly, attention mechanism and TransRalgorithm are added?to complete the log domain knowledge map,the recommendation algorithm is combined with the knowledge graph byconnecting mechanism. Finally, the reservoir parameter information in logging field is predicted and recommended according to thepotential characteristics. Experimental results on logging data sets show that when K = 15 in Top-K recommendation,the accuracy,recallrate and normalized loss cumulative gain of the proposed algorithm are 6. 21% , 8. 05% and 4. 82% higher than those of similaralgorithms respectively. And?it can predict and recommend reservoir parameter information in logging field with high precision and judge reservoir oil and gas condition more accurately,thus revealing that the research scheme of constructing logging reservoir recommendationalgorithm based on knowledge graph is feasible.

摘要：针对实际应用中测井解释人员由于经验不足以及测井解释模型所选取的参数不当、差错等,所造成的测井处理解释结果的评价精度不准等问题,提出了一种基于知识图谱的测井储层推荐算法。 首先,构建测井领域知识图谱对测井解释操作人员和其测井储层特性信息进行统一表示;其次,通过加入注意力机制和 TransR 算法对测井领域知识图谱进行补全,再通过连接机制把推荐算法与知识图谱结合起来;最后,根据潜在特征预测并推荐测井领域中储层参数信息。 在测井领域数据集上的实验结果表明,该推荐算法在 Top-K 推荐中 K = 15 时,其准确率、召回率和归一化折损累计增益三个指标比同类算法分别提高了 6. 21% 、8. 05% 、4. 82% ,并能够高精度预测和推荐测井领域储层参数信息以及更准确地判断储层油气状况,从而揭示出基于知识图谱构建测井储层推荐算法的研究方案是可行的。

关键词：知识图谱;知识图谱补全;测井储层信息;注意力机制;推荐算法