Well-seismic joint multi-target simultaneous inversion method based on state space model and support vector regression

Abstract

The invention relates to a well-to-seismic joint multi-target simultaneous inversion method based on a state space model and support vector regression, and the method comprises the following steps: building a state space model and a state transition transformation matrix based on the actual measurement data of a sample oil well by adopting a discretization method, building a state residual random variable probability density function, establishing an observation function based on a support vector regression method, then establishing an observation error random variable probability density function, and finally performing multi-target simultaneous inversion based on a state space model. According to the method, the difficulty of establishing the state space model is greatly reduced, manpower, material resources and time are saved, and the problems of long period, high cost and low production efficiency of a conventional inversion method are solved.

Description

Technical field: [0001] The invention relates to the field of seismic data inversion and interpretation, in particular, relates to a simultaneous well-seismic multi-object inversion method. Background technique: [0002] Currently, the stochastic inversion method based on logging constraints is the mainstream seismic inversion method. To highlight the characteristics of the stochastic inversion method, researchers and engineers often compare deterministic inversion with stochastic inversion. Deterministic inversion assumes that the wave impedance is a definite value in space, usually based on the convolution model, using the minimization criterion to solve the problem, and obtain a smooth wave impedance estimate. Since the seismic data is band-limited, the biggest limitation of deterministic inversion is that the inversion results lack both low-frequency and high-frequency components. The low-frequency components are usually detected by prestack time / depth migration velocity...

AI Technical Summary

Problems solved by technology

A lot of experience shows that the deterministic inversion results change naturally in the horizontal direction, but the vertical resolution is low, and the single sand body does not match well with the drilling results, which cannot meet the actual production requirements; the random inversion based on logging constraints has a large vertical resolution. The boundary of the thin sandstone is relatively clear, which is basically consistent with the sandstone interpreted by drilling. This is because the initial wave impedance model contains high-frequency logging information beyond the seismic frequency band and has been retained in the inversion results. This high-frequency The frequency information corresponds to the response of the thin layer. Since the distribution of thin layers is usually small and the lateral phase change is fast, there must be enough wells to ensure the reliability of the interpolation results. Higher requirements are placed on the quantity and spatial distribution of logging data, which is more suitable for seismic reservoir description in the development and production stages

Images