The overseas G offshore oilfield has obtained high-quality time-lapse seismic data with the help of OBN technology, and it has been confirmed that there is a correlation between the time-lapse seismic differences in the target layer and the WFCI. However, due to the lack of secondary logging data, it is difficult to establish a quantitative relationship model between the two, which seriously hinders the efficient application of time-lapse seismic differences in oilfield development. This study focuses on solving this problem. An in-depth research is carried out based on the existing data of G oilfield. The core steps are as follows: select an appropriate calculation method for dry rock modulus, construct a typical model and extract sensitive attributes, standardize sensitive amplitude attributes, build, train and apply a deep learning network model, and finally form a deep learning time-lapse seismic prediction method for the WFCI in areas without secondary logging. Forward simulation and practical application in G oilfield show that the WFCI predicted by this method is highly consistent with the actual situation and basically in line with the dynamic understanding, fully demonstrating its good application effect, great reference value for oilfield development and promotion potential. Different from conventional methods, this study generates diverse samples through forward simulation and uses a deep learning network to establish relationships, successfully breaking through the difficulties of prediction and low accuracy in areas without secondary logging, and opening up new ideas and providing new methods for similar research and practice.