Research And Application Of Non-linear AVO Inversion

With the development of oil and natural gas exploration, subtle reservoir has become an important target for the exploration and development of oil and gas field.Lithological reservoirs cover a large portion of subtle traps so, the main goal of this study is to delineate the lithological traps. Since elastic parameters play an important role in the reservoir prediction so, demand for the prediction of elastic parameters have been increased in the petroleum industry. To accurately predict the elastic parameters, this thesis compares and analyzes Zoeppritz’s equations and its approximate formula. In addition, it presents a new non-linear AVO inversion method,which is based on the study of normal AVO inversion methods.In the research of regular AVO inversion methods, this thesis comparatively present several methods including exact Zoeppritz’s equations generalized linear inversion, AVO approximate formula for linear inversion and PP-PS joint inversion.Firstly, the exact Zoeppritz’s equations generalized linear inversion was performed to accurately calculate elastic parameters, but its coefficient matrix condition number for objective function is too large to stabilize the solutions of these equations. However,AVO approximate formula for linear inversion provides more stable solution, but its precision is too low in some situations, such as the large offset and violent variation of elastic parameters. So, PP-PS joint inversion method was utilized, which is based on Fatti approximate formula. It can avoid the Smith-Gidlow approximate formula’s which is strongly dependent on Gardner empirical formula. In addition, it can improve the abilities to identify the fluid component, lithology and reduce the uncertainty of the reservoirs parameters by combining the P and S waves.The AVO inversion is a completely non-linear problem, so we cannot obtain the real solution through linear inversion method. In order to solve the AVO nonlinear problem, this thesis presents a new non-linear AVO inversion method based on Least Square Support Vector Machine(LSSVM) and Particle Swarm Optimization(PSO).LSSVM parameters are optimized through PSO method. This method combines theZoeppritz’s equations, and establishes the objective function by using the LSSVM. In addition, it can directly extract the elastic parameters which have high precision from real seismic records.On the basis of comparative study of several AVO inversion methods, the PP-PS joint inversion and non-linear AVO inversion methods are applied to the real data of the Y area in Pearl River Mouth Basin. The inversion results show that the non-linear method can overcome the problems of regular linear AVO inversion, which occur due to far offset or violent variation of elastic parameters. This method also provides high speed and good stability.