神经网络预测储层砂岩粒度纵向剖面

国内外多年的研究认为，储层粒度特征值（d₅₀，UC）是防砂设计的基础。利用伽马测井、密度测井与粒度特征值之间的相关性，建立探井伽马、密度测井项与实测粒度特征值三者的样本库；利用神经网络技术，训练出满足工程需要的学习网络，进而结合开发区块的测井资料，建立整个储层的粒度纵向分布剖面。该技术对实测数据少或缺乏实测数据储层的防砂显得尤为重要，为防砂方案设计提供了准确的依据，并在现场进行了很好的应用。

Research scholars believe that the characteristic values (d50,UC) of reservoir grain size is the design basis for sand control completion. Based on the relevance of gamma logging, density logging to the grain size characteristic values, we established the sample library of wells gamma, density log and grain size values. Using neural network technology, the learning network project was trained, then combined with logging data development block, the vertical profiles of entire reservoir grain size was established. This technology is very important to sand control completion for the reservoir lack of grain size data. It provides an accurate basis for project design and has a good application effect in the field.