Pre-stack three-parameter inversion implementation reservoir stratum and fluid prediction method based on mixed norm regularization

Abstract

The invention relates to a pre-stack three-parameter inversion implementation reservoir stratum and fluid prediction method based on mixed norm regularization. Based on the Bayesian parameter estimation theory, the AVO basic principle, the comprehensive sparse theory, remaining residual distribution, geological significance and other factors, a mixed penalty functional equation is built, and regularization constraint is performed on the pre-stack three-parameter inversion process. A model and actual material application prove that result stability and geological significance can be improved through mixed norm regularization, and particularly, the density inversion result is more accurate under the situation of the same offset distance and the same signal-to-noise ratio. The method has the advantages of being capable of being directly applied to actual materials, a reasonable inversion result is obtained, and the reservoir stratum and fluid prediction accuracy is improved.

Description

technical field [0001] The invention belongs to the field of petroleum exploration, and relates to a reservoir and fluid prediction method realized by pre-stack three-parameter inversion based on mixed norm regularization, and an elastic parameter inversion method suitable for actual ill-posed AVO (Amplitude versus offset) problems . Background technique [0002] Prestack inversion can simultaneously obtain multiple elastic parameters such as P-wave, S-wave, P-to-S wave velocity ratio, Poisson’s ratio, and Lamé constant, so that a more accurate comprehensive prediction of reservoir and fluid can be made from the elastic domain (Zhang et al. ,2013). The P-wave AVO reflection coefficient equation of the actual continuous multi-interface is approximately expressed as (Buland and More, 2003): [0003] R pp ( t , θ ) ≈ sec ...

AI Technical Summary

Problems solved by technology

[0018] Although the introduction of the prior distribution P(m) can improve the stability of the inversion, usually the assumption that the characteristics of the disturbance variation of the parameters P-wave impedance, S-wave impedance and density are normally distributed ignores the three relationship between

However, the AVO three-parameter inversion problem itself is highly ill-conditioned, and its inversion results (especially the density term) are often extremely unstable

Predecessors have tried to convert the three-parameter equation and remove the density term in the equation to stabilize the inversion, but it is inevitable to balance the contradiction between improving the stability of the inverse problem and the degree of loss of information

In addition, when Theune (2010) compared the inversion results under the assumptions of different prior distributions, he found that artificial horizontal axis artifacts sometimes appeared in the inversion of Cauchy constraints

In fact, the regularization constraints of the existing actual AVO inversion problems generally only focus on the initial model and the degree of agreement with the seismic data, etc., and do not consider the geological significance of the actual elastic parameters, so the final inversion results are often controlled by the initial model. The inherent limitation of the ill-posed problem is that the inversion of elastic parameters such as P-wave, S-wave and density are of different quality and insufficient correlation, which greatly restricts the application of pre-stack inversion in reservoir and fluid prediction

Images