摘要
The development of machine learning and deep learning algorithms as well as the improvement ofhardware arithmetic power provide a rare opportunity for logging big data private cloud.With the deepeningof exploration and development and the requirements of low-carbon development,the focus of exploration anddevelopment in the oil and gas industry is gradually shifting to the exploration and development of renewableenergy sources such as deep sea,deep earth and geothermal energy.The traditional petrophysical evaluation andinterpretation model has encountered great challenges in the face of new evaluation objects.To establish a distributedlogging big data private cloud platform with a unified learning model as the key,which realizes the distributed storageand processing of logging big data,and enables the learning of brand-new knowledge patterns from multi-attributedata in the large function space in the unified logging learning model integrating the expert knowledge and the datamodel,so as to solve the problem of geoengineering evaluation of geothermal fields.Based on the research ideaof“logging big data cloud platform---unified logging learning model---large function space---knowledge learning&discovery---application”,the theoretical foundation of unified learning model,cloud platform architecture,datastorage and learning algorithm,arithmetic power allocation and platform monitoring,platform stability,data security,etc.have been carried on analysis.The designed logging big data cloud platform realizes parallel distributed storageand processing of data and learning algorithms.New knowledge of geothermal evaluation is found in a large functionspace and applied to Geo-engineering evaluation of geothermal fields.The examples show its good application in theselection of logging series in geothermal fields,quality control of logging data,identification of complex lithologyin geothermal fields,evaluation of reservoir fluids,checking of associated helium,evaluation of cementing quality,evaluation of well-side fractures,and evaluation of geothermal water recharge under the remote logging module ofthe cloud platform.The first and second cementing surfaces of cemented wells in geothermal fields were evaluated,as well as the development of well-side distal fractures,fracture extension orientation.According to the well-sidefracture communication to form a good fluid pathway and large flow rate and long flow diameter of the thermalstorage fi ssure system,the design is conducive to the design of the recharge program of geothermal water.
机器学习与深度学习算法的发展及硬件算力性能的提升,为测井大数据私有云的构建提供了难得的机遇。随着油气勘探开发程度的不断加深及低碳发展要求的提出,油气行业勘探开发重心正逐步向深海、深层、地热等可再生能源领域转移。面对新型评价对象,传统岩石物理评价解释模型面临巨大挑战。本文旨在构建以统一学习模型为核心的分布式测井大数据私有云平台,实现测井大数据的分布式存储与处理;并在融合专家知识与数据模型的统一测井学习框架中,通过大函数空间下的多属性数据挖掘,实现对全新知识模式的学习,进而解决地热田地学工程评价难题。基于“测井大数据云平台—统一测井学习模型—大函数空间—知识学习与发现—工程应用”的研究思路,本文系统分析了统一学习模型的理论基础、云平台架构设计、数据存储与学习算法、算力调度与平台监控、系统稳定性及数据安全等关键技术问题。所设计的测井大数据云平台实现了数据与学习算法的并行分布式存储及处理,通过大函数空间挖掘获得地热评价新知识,并将其应用于地热田地学工程评价实践。实例验证表明,该平台在云平台远程测井模块支持下,在地热田测井系列优选、测井数据质量控制、复杂岩性识别、储层流体评价、伴生氦气检测、固井质量评价、井旁裂缝评价及地热回灌方案设计等场景中均表现出良好的应用效果。应用该平台完成了地热田固井井一、二界面固井质量评价,以及井旁远端裂缝发育特征、裂缝延伸方向的识别;基于井旁裂缝连通形成有效流体通道及热储裂隙系统大流量、长流径的发育特征,为地热回灌方案优化设计提供了技术支撑。
作者
Cheng Xi
Fu Hai-cheng
He Jun
程希;傅海成;何军(西安石油大学地球科学与工程学院,陕西西安710065;油气藏地质及开发工程国家重点实验室(西南石油大学),四川成都610500;西安石油大学院士专家工作站,陕西西安710065;中国石油勘探开发研究院;中国石油天然气股份有限公司天然气销售陕西分公司,陕西西安710000)
基金
supported by Grant(PLN2022-14)。
