The Study Andapplication Of The Multiscale Seismic Data Joint Inversion Method

With the development of oil and gas exploration situation, hydrocarbon exploration is gradually shifted from structural reservoir to lithologic reservoirs. As a most effective geophysics method for reservoir exploration, seismic technology has been shifted from structure characterization to reservoir evaluation, lithologic interpretation and reservoir characterization and so on. Seismic inversion can extract various information about lithology, physical properties and fluid from seismic data, to help provide rich information for reservoir description. However, with increasing difficulty of exploration, conventional seismic inversion has failed to provide enough useful information. In order to perform high-precision exploration, I integrate the surface seismic data, crosswell and VSP data in the inversion method, making full use of the high lateral resolution of surface seismic data, wide-band information of crosswell data and accurate time-depth relationship of VSP data, to improve the accuracy of seismic inversion, and provide better service for exploration.Conventional impedance inversion only uses the surface seismic data. Although surface seismic data has high lateral resolution, its vertical resolution is low. Surface seismic data, crosswell seismic data, and VSP data, although acquired in different means, of different resolution, are the corresponding seismic wave responses of the same geologic body. With the help of the Bayesian theory, we can link these three data together, and set up the joint probability density function of them.In this thesis, I first discussed the Bayesian Theory, and studied multiscale seismic data joint inversion method. Then, I studied the influence of different prior distribution on the inversion results, and finally chose the Cauchy distribution, which fits the real well data well. Besides, I studied the influence of different parameters on the inversion results. In solving the inversion equation, I used the modified PRP conjugate gradient method, accelerating the convergence rate and increasing the robustness of the method. I also studied the sequential Gaussian simulation method for modeling. This method avoids the smoothing effect of the determinable modeling, and can help optimize the best results, as well as assessing the uncertainty. Finally, considering that real data has giant difference in value, I preprocessed the real data, making the choice of inversion parameters easier. Multiscale seismic data joint inversion method is tested with theoretical model, by comparing the results with the conventional inversion, it shows that the result of joint inversion not only has better lateral continuity, but also has higher resolution. Through the test of real survey data, multiscale seismic data joint inversion reflects higher resolution, and its overall performance is better than the conventional inversion method.