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。 Predicting reservoirs using pattern recognition can diagnose the type of damage and the degree of damage, which provides an important basis for the optimization design of drilling fluids, completion fluids and other working fluids. By weighting the conventional Euclidean distance, the core problem of pattern recognition is solved - membership function is established, and a new model of predicting reservoir sensitivity by using pattern recognition is established and successfully applied. Taking water sensitivity as an example, the eigenvectors were determined by feature selection and extraction, and the average model of water damage was established based on the classification of damage degree. With 127 sets of data from Dagang Oilfield, the new model was applied to predict the reservoir sensitivity . The results show that the average accuracy rate of water sensitivity index prediction is more than 86.9% and the prediction success rate of water damage degree reaches 90.0%. It is proved that the new model predicting reservoir sensitivity using pattern recognition method has high accuracy of prediction results. Reliable and other advantages, to improve the protection of oil and gas reservoir and plugging effect is of great significance.