摘要：

随着油田能源系统的复杂性不断增加，传统的监控、分析和优化的方法往往难以应对来自不同来源的大量数据，导致在识别和解决能耗异常方面效率低下，难以实现能源利用的最佳效果。为克服这些局限性，实现油田集输与注水系统能源管控的智能决策，本研究针对海量多源异构数据，提出一种基于知识图谱的能耗异常智能辅助决策方法。以能耗异常台账及操作手册等文本资料为主要数据源，建立能源管控知识内容体系框架，作为组织和整合多源数据的基础，确保数据的高效利用。采用BiGRU-CRF对文本资料进行实体抽取，识别设备、参数、异常等关键概念；采用BiGRU-ATT进行实体间关系抽取，捕捉油田集输与注水系统中复杂的相互依赖性；通过Neo4j图数据库对提取到的能耗知识进行存储和可视化展示，其结构化的知识表示形式为后续数据的高效利用奠定了基础。根据所构建的知识图谱开发能源管控可视化平台，提供用户友好的界面，使操作人员能够以直观的方式探索能耗知识。该平台从数据和知识层面提供可行的措施推荐，以辅助指导能耗控制。油田现场的应用结果表明，采用所提出的基于知识图谱的油田集输与注水系统能耗异常智能辅助决策方法，能够融合多源异构数据，为工艺流程中发生的能耗异常事件提供及时性、整体性、智能性的辅助决策推荐，指导操作人员进行快速有效的能耗控制，显著减少了决策所需的时间。本研究为油田能源管控建设提供了新的思路，对其它油田的能耗控制管理具有指导意义。

关键词: 集输系统, 注水系统, 能耗异常, 知识图谱, 辅助决策

Abstract:

The increasing complexity of energy systems in oilfields necessitates advanced approaches to monitor, analyze, and optimize energy usage. Traditional methods are often inadequate for processing the vast amounts of data generated from diverse sources, leading to inefficiencies in identifying and resolving energy consumption anomalies and making it difficult to achieve optimal energy utilization. To overcome these limitations and achieve the intelligent decision-making for energy management and control in oilfield gathering and water injection systems, an intelligent assisted decision-making method for abnormal energy consumption was proposed based on knowledge graph, addressing the challenges posed by massive multi-source heterogeneous data. Specifically, the abnormal energy consumption records and operation manuals were utilized as the primary data source, and the comprehensive knowledge framework for energy management and control was established. This framework serves as the foundation for organizing and integrating multi-source data, ensuring systematic and efficient data utilization. Additionally, the BiGRU-CRF (Bidirectional Gated Recurrent Unit-Conditional Random Field) model was applied to extract entities from the textual data, identifying key concepts such as equipment, parameters, and anomalies. And the BiGRU-ATT (Bidirectional Gated Recurrent Unit-Attention) model was adopted to extract relationships between entities, capturing the complex interdependencies within the oilfield gathering and injection systems. The extracted energy consumption knowledge is stored and visualized using the Neo4j graph database, providing a robust platform for data querying and analysis. Its structured representation lays the foundation for the efficient utilization of data in subsequent stages. Finally, based on the constructed knowledge graph, an energy management and control visualization platform was developed, providing a user-friendly interface that enables operators to explore energy consumption data and knowledge in an intuitive manner, significantly enhancing the usability of the operational system. The platform provides actionable recommendations at both the data and knowledge levels, supporting energy consumption control effectively. The field application results in oilfields demonstrate that the proposed intelligent decision-making method, based on knowledge graphs, effectively integrates multi-source heterogeneous data for abnormal energy consumption detection in oilfield gathering and injection systems. Timely, comprehensive, and intelligent decision-making recommendations are provided for energy consumption anomaly events in the gathering and injection processes, guiding operators in achieving rapid and effective energy consumption control. The time required for decision-making is significantly reduced through this method. This study offers a novel and impactful approach for the construction of energy management and control systems in oilfields, which provides valuable guidance for the management of abnormal energy consumption in other oilfields.

Key words: gathering and transportation system, water injection system, abnormal energy consumption, knowledge graph, assisted decision-making

中图分类号: 

• TE866

• TE863