应用测井和BP神经网络算法预测储层敏感性

Using Logging and BP Nerve Net Calculation Methods to Predetermine Sensitivity of Reservoirs

在收集薄片、铸体薄片、粒度、压汞、X—衍射、扫描电镜、物性、敏感性流动实验等岩心分析资料的基础上 ,首先通过单相关分析找出影响敏感性的主要因素 ,然后再应用测井资料提取这些敏感性参数 ,最后以影响敏感性的主要因素做为 BP神经网络的输入层 ,应用 BP神经网络算法 ,建立敏感性预测模型 ,预测储层敏感性。该方法对西部某油田的资料进行了处理分析。结果表明 ,速敏、水敏、盐敏、酸敏和碱敏的预测结果与该油田的敏感性流动实验结果的符合率为 80

Based on collecting data from the thin slice observation, cast slice observation, granularity analysis, mercury injection, micro scanning, porosity and permeability analysis, X-ray diffraction, core fluid sensitivity tests etc., the main factors affecting reservoir sensitivity are obtained with single variable regression method. These data extracted from well logging information are used to train the BP neural network to predict reservoir sensitivity. This method is applied to process and analyze the data from oilfield. The results show that the coincidence rate comparing the predication results of fluid velocity sensitivity, fresh water sensitivity, salt-water sensitivity, acid sensitivity and alkali sensitivity in this oilfield is about 80%. Obviously, this method is effective and practical.