Deep Learning Wave Field Characteristic Analysis And Prediction Technology Of Reservoir

There are abundant fracture-vuggy reservoirs in Xinjiang,where a large number of reservoirs are deeply buried.The deep geological conditions are much more complicated than the middle and shallow ones,which result in weak seismic reflection signal,low signal-to-noise ratio and resolution of data,poor quality of seismic data,complex seismic wavefields and difficult seismic imaging,etc.These factors cause great difficulties in reservoir identification.The fracture-vuggy reservoirs can be used as both migration and storage of oil and gas.Fracture-vuggy reservoirs have many kinds of storage spaces,including pores,karst caves and fractures,which usually exist underground as fracture-vuggy system.In order to serve reservoir exploitation,it is of great significance to study seismic wavefield response of fractures and vugs.In this paper,the wavefield characteristics of fault-karst and intelligent identification of fracture-vuggy zones are studied,aiming at establishing the corresponding relationship between different geological models and their wavefields,so as to promote the identification of deep fracture-vuggy reservoirs.This paper uses random medium model method to establish fracture velocity models,and adds the structural setting,in turn,fault-karst geological models are set up.By seismic wave forward numerical simulation,analyzing the wave field characteristics of the fault-karst with different geological features,the corresponding relation of wave field characteristics and geological model is established.In terms of artificial neural network,a large number of velocity models of fracture-vuggy zones are established by random medium modeling method to obtain training data.Finally,through the training of network,the network can be used to synthetic and real seismic data of the Tahe oilfield for intelligent identification of fractures and vugs.