摘要
针对具有复杂储集空间和极强的非均质性的低孔低渗储层,常规测井响应特征不够明显,使用传统解释手段难以有效识别储层流体的问题,提出了一种基于小波变换和CNN-Transformer混合模型的储层流体识别方法。首先,使用小波变换将测井信号从时域扩展到时频域,并生成时频谱图以增强信号特征,然后使用滑动时窗沿着测井曲线深度方向滑动采样,获取代表解释深度处地层信息的频谱特征图,最后,通过训练CNN-transformer模型深度挖掘特征图信息,实现储层流体识别。混合模型在利用储层对应深度处测井数据的同时,又兼顾测井曲线随深度的变化趋势和地层前后信息的关联性,挖掘时频谱图的局部细节和全局特征表示,自动识别流体类型。将模型应用于大港油田22口实测测井资料中,并与CNN和BiLSTM等多个模型的流体识别效果进行对比分析,基于小波变换和CNN-Transformer模型识别效果明显优于其他方法,在测试集上识别准确率达到了92.7%。研究结果表明该方法可以作为低孔渗油藏常规测井资料识别储层流体的有效手段,为流体评价提供了新思路。
Low-porosity low-permeability reservoirs have complex storage spaces and strong heterogeneity,and the characteristics of conventional logging responses are not clear enough,which makes it difficult to effectively identify reservoir fluids using traditional interpretation methods.For this purpose,a hybrid model combining wavelet transform and CNN-Transformer is proposed for reservoir fluid identification.Firstly,the logging signal is extended from time domain to time-frequency domain using wavelet transform,and a time-frequency spectrum is generated to enhance the features of the logging signal.Then,the sliding sampling is carried out along the depth direction of the logging curve by sliding time window to obtain a spectral feature map representing the formation information at interpretation depth.Finally,the information in the spectral feature map is deeply mined by training the CNN transformer model to achieve the identification of reservoir fluid.The model was applied to the logging curves of 22 wells in Dagang Oilfield,and its fluid identification results were compared with those of multiple models such as CNN and BiLSTM.The results showed that the recognition performance of this model was significantly better than other models,with a recognition accuracy of 92.7% on the test set.The research results indicate that this method can be an effective means of identifying reservoir fluids using conventional logging data in low porosity and permeability reservoirs,providing new ideas for fluid evaluation.
作者
龚安
张恒
GONG An;ZHANG Heng(College of Computer Science and Technology,China University of Petroleum(East China),Qingdao,Shandong 266580,China)
出处
《西安石油大学学报（自然科学版）》
CAS
北大核心
2024年第4期108-116,共9页
Journal of Xi’an Shiyou University（Natural Science Edition）
基金
中石油重大科技项目“多级压裂水平井产能评价规律及应用”(ZD2019-183-004)
中央高校基本科研业务费专项资金“人工智能非常规油藏储层评价算法研究”(20CX05019A)。
