阿曼五区块Daleel油田储层裂缝识别方法研究

On Identification Methods for Reservoir Fractures in Daleel Oilfield in Oman Block-5

在对研究区阿曼五区块Daleel油田储层地质特征分析研究的基础上,综合常规测井资料和取心资料构建35个样本。利用常规测井裂缝识别模式对储层裂缝进行识别,将有效裂缝分离出来;引入模式识别领域应用较好的支持向量机(SVM)方法,通过训练样本的分布和实验结果选择核函数类型,利用网格搜索寻优法得到模型最优参数,建立起碳酸盐岩裂缝识别模型。利用该模型对8个预测样本进行识别,正确识别的有7个,预测精度达87.5%,其中1个误判样本是将无效裂缝判识为非裂缝。对27个建模样本进行回判,准确率达100%。通过岩心资料对比发现,当描述一个研究目标与多个相关地质因素的非线性关系强烈的应采用支持向量机(SVM)算法,其计算速度比人工神经网络(ANN)快10倍以上,且判断程度更准确。应用SVM方法,综合考虑储层的岩性、物性和裂缝特征等多种因素建立裂缝识别模型,可以提高裂缝测井解释精度。

Daleel oilfield in Oman Block-5 is a fauh-lithology oil reservoir, which developed in a uniclinal structure higher in the southwest and lower in the northest. The reservoir is biogenicdebris and grained carbonate rock of the upper early Cretaceous. Fracture identification is important in oilfield development. On the basis of detailed analyzing geologic characteristics of Daleel oil field, the conventional logging data was used in combination with the well testing data to select and normalize 35 samples. Firstly, fractures of reservoir are identified with conventional logging data and fracture recognition patterns in Daleel oilfield. Effective fractures are identified and iso lated from others. But the fractures with higher filling degree can not be identified effectively. So the support vector machine in pattern recognition domain is introduced in distinguishing frac tures. Then, the kernel function is determined by referring the distribution of training samples and the experimental comparison. Further more, the optimum parameters are identified by net work searching to establish the fluid recognition patterns for carbonate fractures. The pattern is used to distinguish 8 predicting samples. As a result, 7 samples are correctly distinguished and the accuracy degree is 87.5%. The other ineffective fracture samples are interpreted as non fractures. The 27 samples are distinguished again, with an accuracy degree of 100%. Through the contrast of the cores, it is feasible to correctly distinguish the reservoir fractures by support vector machine method.