论文详情
裂缝识别和评价是裂缝性储层测井解释的核心任务。总结了利用常规测井资料对裂缝性储层进行分级评价的基本方法,为提高效率并减少各种影响因素的干扰,引入卷积神经网络这种深度学习算法进行裂缝识别和储层等级评价。理论分析认为,凭借局部权值共享的特殊结构,卷积神经网络比现有其他算法在提取数据特征时具有更显著的优越性。将裂缝性储层的常规测井数据和储层已知信息作为标准样本输入卷积神经网络进行学习,然后对塔里木盆地北部某含油气构造单元其它井的有关层段进行裂缝识别和储层评价,结果表明,卷积神经网络的评价结果比传统方法更高效、准确,可以根据测井数据直接输出裂缝的发育等级。
The core task of logging interpretation in fractured reservoirs is to identify and evaluate fractures.Basic methods that use conventional logging data to classify fractured reservoirs are summarized.Then,to increase the efficiency and reduce the interference of various influencing factors,a deep-learning algorithm,the convolutional neural network (CNN),is introduced to identify the fractures and evaluate the reservoirs in this research.Theoretical analysis indicates that,because of the special structure of local weight sharing,the CNN algorithm possesses more significant advantages than other algorithms in extracting data features.The conventional logging data and known reservoir information of a fractured reservoir in some wells of the north Tarim Basin were used as standard samples to enter into the CNN model for learning,and then the relevant intervals of other wells in the region were identified and evaluated using this model.The evaluation result of using CNN technology was more efficient and accurate than that of the traditional method,and it can directly output the development level of cracks based on the logging data.
国家自然科学基金项目(40874057)资助。