超深井储层潜在敏感性预测

Prediction of Reservoir Potential Sensitivity on Ultra-Deep Wells

从现有储层岩性、物性资料以及前期积累的试验数据中,找出了造成超深井储层敏感性损害的各种潜在因素,并进行了归一化和定量化处理。利用Matlab数学计算软件的神经网络工具,建立了各种潜在损害因素与储层敏感性伤害之间的神经网络模型,并利用各种潜在损害因素归一化和定量化处理的结果,对网络进行了训练,利用返回检验法验证了该神经网络模型预测储层敏感性损害的准确率,准确率在85%以上。最后利用该神经网络模型对胜科1井深部储层敏感性进行了预测。

From reservoir lithology, reservoir properties and other experimental data, various potential factors, which might damage reservoirs, were identified, normalized and quantified. Based on these data, a BP network is established using MATLAB neural network tool and is trained. The comparison of the simulation result with that of traditional regression method shows that the prediction result of neural network is more precise and up to 85% . The formation sensitivity was predicted in well Shengke-1.