The Constrat Study Of Pre-stack Seismic Inversion And Reservoir Prediction

Post stack seismic inversion in reservoir prediction is the core,which is mainly based on seismic,logging and geological data,obtain the wave impedance data or attributes data that can reflect the spatial distribution characteristics of underground sand,and ensure the unity of the lateral continuity and vertical resolution of the data,which not only can achieve the vertical resolution of sandstone,at the same time,to meet the real geological law,provide the basis for the prediction of favorable reservoirs.According to the characteristics of reservoir thickness,sandstone distribution is complex between well and well,well unequal distribution in the Sha2 Member of Chengbei low step-fault zone of southern Qikou Sag,in turn including contrast study of constrained sparse spike inversion,Bayesian stochastic inversion and neural network multi-attribute inversion,to predict reservoir.In the early stage of the inversion,standardizing and sensitivity analysis of 20 wells in the study area,by analysis of the sensitive parameters,optimization of gamma and sonic curve to reconstruct pseudo sonic curve,improve curve recognition accuracy in sandstone.By constraining sparse spike inversion and fully applying seismic data,the spatial distribution of sandstone is initially determined.Bayesian stochastic simulation inversion fusion constraint sparse spike inversion data,using Markov chain-Monte Carlo simulation algorithm to improve the longitudinal resolution while reducing reservoir uncertainty;Neural network multi-attribute inversion,based on seismic attribute and lithological characteristic curve,using neural network algorithm to establish the nonlinear relationship between attribute and lithological parameters,using the high resolution of the logging data and the overall nature of the attribute,to solve the problem of uneven distribution of wells,accurate prediction of sandstone distribution.Through the analysis of the inversion results,the constrained sparse spike inversion can be better recognized for the thick sandstone,but the resolution of the sand body below 10 m can not be achieved.Bayesian stochastic simulation inversion resolution is high,can identify about 5m thin sandstone,but in the well or no well area,sandstone prediction accuracy is low;Neural network multi-attribute inversion,the vertical resolution can be up to 5m,and the use of attribute constraints,in the absence of wells and less wells in the sandstone prediction accuracy is higher.So the final use GR data body which from neural network inversion to track and depict sandstone,combined with drilling test oil data to predict favorable reservoir distribution area.Based on the fine sand depiction,12 sets of main sandstone with good oil content and large distribution range are depicted,and 36 sets of non-main sandstone with relatively poor oil content and limited sandstone distribution area are depicted.The analysis of the main and non-main sandstone was carried out to determine the favorable area for the well Chenghai 33 and well Chenghai 37,the favorable sections in the longitudinal direction are Es2z1?Es2z2?Es2x2.