Research Of Multi-point Geostatistical Inversion

The traditional geostatistical inversion improves the prediction accuracy of seismic inversion and has been favored by many scholars in the last 20 years.However,with the deepening of the application of geostatistical inversion,the researchers found that it was plagued by the complex spatial distribution of reservoirs,and the constraints of two geostatistics on the representation of reservoir morphologies and models.The result of the performance is more error in the complex area.It's necessary to develop a new technology to improe accuracy of seismic inversion The multi-point geostatistical method is a modeling method considering sedimentary model developed on the basis of two-point geostatistical method,and it's advantages has been proved to be better than two-point geostatistical.Based on this point,this paper introduces multi-point geostatistical into seismic inversion and develops a new set of multi-point geostatistical seismic inversion method.Firstly,the probabilistic relationship between elastic parameters and lithofacies was established,and the sedimentary facies training images of the study area were determined by geological research.Secondly,the seismic elastic parameters were used to constrain the multiphase modes and the seismic elastic properties were extracted randomly by the types.Thirdly,The obtained seismic elastic properties are convolutioned to form a seismic trace,and compared with the actual seismic trace,the optimal convolution trace is selected as the simulation result and its corresponding elastic attribute is given to the simulated mesh.The above process is repeated until all the meshes are accessed,To complete an iteration to achieve.Finally,whether to repeat the iteration according to the simulated seismic traces and the actual seismic traces until the best fit is reached,and multi-point geostatistical inversion is completed.finally,output the realizationOn the basis of the completion of the algorithm,five 2-D sand-mud models and three 3-D model were built to test the new method.The results shows that multi-point geostatistical inversion can effectively identify sand and mudstone interlayer,as well as distinguish between different fluid sandstone.The inversion synthetic record section is highly similar to the real seismic data,which can reflect the sandstone and channel,and reproduce the elastic parameter field of different types of reservoir.The research of the training image shows that the seismic data has a constraint effect on the inversion results and the effect of the training image on the inversion results and the convergence speed.Multi-point geostatistical inversion has an advantage in the characterization of reservoir and the reduction of channel.The ability to identify sandstone is about 4m.There is a lot of space for the study of multi-point geostatistics inversion to improve.