Research And Application Of Tight Sandstone Reservoir Prediction Method Based On Deep Learning

Tight sandstone oil and gas resources account for a large proportion in unconventional oil and gas resources,but there is still a lack of relatively complete technical means in tight sandstone reservoir prediction.Traditional methods lack both applicability and accuracy.Lishu area is located in Songliao Basin.Certain exploration achievements have been made in this area,but the next exploration target is not clear,so it is necessary to further forecast the tight sandstone reservoir.Based on the existing seismic data and well logging data in the study area,this paper optimizes the seismic data by means of hyperchannel set,and supplements the well logging data by means of shear wave velocity prediction and standardization to eliminate errors.Based on the petrophysical analysis,it is optimized that the sensitive parameters of lithology are density and shear modulus,and the sensitive parameters of porosity are p-wave impedance,density and sandstone content.The density,shear modulus and measured depth logging curves are taken as input,and the actual logging content of sandstone is taken as label.A prediction model of sandstone content is constructed by using the fully connected deep neural network.The density,p-wave impedance and actual sandstone content logs were taken as input,and the actual porosity logs were taken as labels.The full-connection depth neural network was used to construct the porosity fitting relationship.The results of prestack inversion were put into the lithology and porosity prediction model trained by the fully connected deep neural network to predict the distribution of high-quality reservoirs in the study area.Young’s modulus and Poisson’s ratio were used to calculate the brittleness index and obtain the brittleness parameters in the study area.The coordinate rotation technique combined with fluid sensitive parameters,p-wave impedance and p-wave velocity ratio was used to realize fluid prediction in the study area.Based on the above prediction results,the distribution of tight sandstone "sweet spot" in the study area is analyzed.In the study area,the "sweet spot" of tight sandstone is mainly located in the southeast and north of the second member of T41 Formation,while the development of K1Y1 Formation is less in the southeast of the upper member of T41 Formation.