Research On The Application Of Seismic Identification And Refined Description Of Low-Grade Faults

Fault identification is one of the important links in the interpretation of seismic data,and the identification of low-grade faults is critical for deploying wells,resolving injection-production contradiction,and improving recovery ratio during oilfield development.This study is based on the Yesanbo Oilfield,starting from the interpretive processing of seismic data,and using deep learning algorithms to achieve high-precision automatic fault identification technology.This paper has proposed three aspects of research work on the detection and identification of low-grade faults.(1)In view of the poor quality of raw seismic data(low signal-to-noise ratio and low resolution),combined with the median filtering technology and diffusion filtering technology based on structure-oriented to perform interpretive processing work on seismic data,and achieve the effect of fault enhancement and enhance the ability of seismic data to reflect low-grade faults.(2)Based on the Tensorflow framework to build the U-net network and the improved Res-Unet network,and the fault identification technology based on deep learning was realized by using the forward synthetic fault data,which improves the efficiency and accuracy of fault interpretation compared with the traditional method.(3)Using the ant-tracking technology in Petrel software to optimize the fault probability volume identified by deep learning,extract the3 D fault fragment and analyze the characteristics of fault elements.Finally,we got the3 D interpretation of the faults.