人工智能在致密储层裂缝测井识别中的应用

Application of Artificial Intelligence in Fracture Identification Using Well Logs in Tight Reservoirs

裂缝是致密储层的有效储集空间和重要渗流通道,裂缝对致密储层的勘探与开发至关重要.单井裂缝识别主要使用井壁成像测井、阵列声波测井和常规测井,如何准确识别裂缝是致密储层研究领域的关键性难题.人工智能是新时期油气智能勘探开发突破现有技术的局限性、提高单井裂缝识别精度的利器.因此,结合近年来人工智能在致密储层裂缝识别的案例及笔者团队在该领域的研究工作,分别介绍了无监督学习、有监督学习和半监督学习人工智能方法在三类测井数据裂缝识别中的应用现状.目前,人工智能在常规测井裂缝识别中应用最为广泛,在井壁成像测井裂缝识别次之,在阵列声波测井识别中应用相对较少.关于人工智能算法,无监督方法由于识别精度问题,应用相对较少;有监督学习方法是目前应用的主流方法,但其需要有充足的有标签数据才能建立有效的裂缝预测模型;半监督学习方法是近年来的新趋势,其可以融合无监督和有监督学习的优点,充分利用有标签测井小样本数据和无标签测井大样本数据,但运行效率是该类方法需要改进的地方.目前单井裂缝人工智能识别方法的发展趋势是往高非线性拟合能力发展、单方法预测往多方法集成发展.同时也系统讨论了各类人工智能方法的存在的问题及未来的发展趋势.

Fractures are effective reservoir spaces and important seepage channels of tight reservoirs.Fractures are very important for the exploration and development of tight reservoirs.Fracture identification of single well mainly can use image logs,array acoustic logs and conventional logs.How to accurately identify fractures is a key problem in the field of tight reservoir research.In the new era of intelligent exploration and development of oil and gas,artificial intelligence is a powerful tool to break through the limitations of existing technology and improve the accuracy of fracture identification in single well.Therefore,combined with the fracture identification cases using artificial intelligence in tight reservoirs in recent years and our research in this field,this paper introduces the application in fracture identification using the three types of logging data by unsupervised learning,supervised learning and semi-supervised learning artificial intelligence methods.So far,artificial intelligence is most widely used in fracture identification using conventional logging,followed by imaging logs,and array acoustic relatively less.As for artificial intelligence algorithms,unsupervised methods are less used because of the problem of recognition accuracy.Supervised learning methods are the mainstream at present,but it needs sufficient labeled data to establish an effective fracture prediction model.Semi-supervised learning method is a new trend in recent years,which can integrate the advantages of unsupervised and supervised learning,and make full use of small sample data of labeled logging and large sample data of unlabeled logging.Noted the operation efficiency of this kind of method needs to be improved.At present,the development trends of fracture identification methods by artificial intelligence for single well are from lower to higher nonlinear fitting ability and integrate multiple single prediction methods into an ensemble prediction method.At the same time,this paper also systematically discusses the existing problems and future development trend of artificial intelligence methods for fracture identification in tight reservoirs.