摘要
针对复杂岩性碳酸盐岩储层孔隙度预测精度不高的问题,首先利用邻域粗糙集选取孔隙度敏感曲线,将选取后的曲线作为BP神经网络的输入,建立BP神经网络孔隙度预测模型;针对BP神经网络存在的易陷入局部最小等问题,对网络拓扑结构参数进行调整实验,用磷虾觅食算法对BP神经网络初始权值阈值优化,建立基于粗糙集-磷虾觅食算法的神经网络孔隙度预测模型。将此模型应用于某油田A井中,与改进前BP神经网络、体积模型等模型精度对比,证明了新网络模型的优越性。结果表明新模型回判将预测精度从26.5%减小到7.05%,具有更强的学习能力,更适用于复杂岩性储层总孔隙度的测井评价。
In view of low porosity prediction accuracy in the carbonate reservoir with complex lithology,firstly,the neighborhood rough sets are applied to selecting porosity sensitivity curves.On the basis of rough set selected curves,the BP neural network porosity prediction model is established.The network topology parameters are adjusted and the initial weights and thresholds of BP neural network are optimized by krill foraging algorithm in order to address the issue that the BP neural network is susceptible to the occurrence of local minima,a neural network porosity prediction model which is optimized by krill foraging algorithm and rough sets is proposed.The model was applied to A well of an oil field.The comparison between the new model and the former BP neural network,volume model and other models shows that the improved porosity evaluation method can improve the evaluation accuracy of reservoir porosity.The result shows that the relative regressional error of the new modelis reduced from 26.5% to 7.05%.From the above,we learn that the new model has better learning abilityand generalization ability.The new model is more suitable for logging evaluation of porosity in the carbonate reservoir with complex lithology.
出处
《中国科技论文》
北大核心
2017年第9期990-998,共9页
China Sciencepaper
基金
国家自然科学基金资助项目(41504094)
