模式识别在储层敏感性预测中的应用  

Application of pattern recognition in the prediction of reservoir sensitivity

作　　者：蒋官澄[1] 王晓军[2] 吴雄军[2] 张卫行[2] 胡成亮[3] 熊英[3] 何丕详[3] 

机构地区：[1]中国石油大学(北京)石油与天然气工程学院,北京102249 [2]中国石油大学(华东)石油工程学院,山东东营257061 [3]中国石油大港油田分公司采油工艺研究院,天津300280

出　　处：《油气地质与采收率》2010年第5期61-64,共4页Petroleum Geology and Recovery Efficiency

基　　金：国家科技重大专项"复杂结构井储层损害评价与保护技术"(2009ZX05009-005);国家杰出青年科学基金"油气层损害与保护"(50925414)

摘　　要：利用模式识别对储层进行敏感性预测可以对其损害类型及损害程度进行科学诊断,从而为后续钻井液和完井液及其他工作液的优化设计提供重要依据。通过将常规的欧氏距离进行加权改进,解决了应用模式识别的核心问题——构建隶属函数,进而建立了采用模式识别法预测储层敏感性的新模型,并得到了成功应用。以水敏为例,经过特征选择与提取确定特征向量,利用损害程度等级的划分建立水敏损害的均值样板,借助大港油区127组数据检验新了模型在储层敏感性预测中的应用效果。结果表明,水敏指数预测的平均准确率大于86.9%,水敏损害程度的预测成功率也达到了90.0%,证明采用模式识别法预测储层敏感性的新模型具有预测结果准确性高、结论可靠等优点,对提高油气层保护和油气层解堵效果具有十分重要的意义。Rapid and accurate prediction of reservoir sensitivity are helpful to establish the scientific base in the designing of oil field development plan, and it has become a hot research point for scholars home and abroad. The rapid development of pattern recognition tech-nology makes the prediction of oil reservoir sensitivity feasible. In this paper,by means of transforming conventional euclidean distance to weighted euclidean distance, we can solve the core issues that is the most critical to the construction of membership functions , then a pattern recognition prediction model is established, moreover, the new model was applied for the prediction of sensitivity successfully and proved by satisfactory results. We determine the feature vectors of various damage types at first,and then the performance of new model for the reservoir sensitivity prediction is tested by 127 sets of collected data. This method is feasible that the prediction accuracy rate is greater than or equal to 86.9% ,and the prediction success rate of water-sensitive damage level also reaches 90.0%,which show that the new pattern recognition model can predict the reservoir sensitivity accurately and reliably. So, the new forecasting model will play an important role in reservoir protection and plugging.

关 键 词：模式识别 储层 敏感性预测 特征选择 隶属函数 

分 类 号：TE353.3[石油与天然气工程—油气田开发工程]