Estimation of Regular Parameters for Impedance Inversion

Abstract

Using the regularization, we can turn an ill‐posed inverse problem into a good‐posed one. Through selecting a reasonable initial value of regular parameters by the statistic method and estimating regular parameter values by the maximum likelihood (ML) in inverse procedure, the inverse result and convergent speed can be improved. Meanwhile, we combine the regularization and the fast simulated annealing algorithms to play the full role of regular parameters, then we can acquire a global optimum result and gain good applied effect.

Adaptive reconstruction method of impedance model with absolute and relative constraints

“…The smooth regions have small impedance gradients, and the edges or discontinuities have large gradients. If δ value is over-small and constant to the whole model, the variable D c k (Z)/δ is enlarged, and the small gradients D c k (Z) are easily accepted so that the results at the edges are over-smoothed (Zhang et al, 2005). On the other side, if δ values are over-large, the large gradients are easily accepted so that the results in the smooth regions are unstable.…”

“…N 1 and N 2 values are obtained according to the impedance gradients at the layer interfaces and the fault zones, respectively. Note that the non-uniform δ used in the whole model is different from adaptively adjusting δ in the inversion procedure (Sen and Roy, 2003;Zhang et al, 2005).…”

“…So we want to rebalance J 1 with J 2 or J 3 by adjusting the regularization parameters in the inversion procedure. The maximum likelihood method is an effective method of adjusting the regularization parameters in the inversion procedure (Zhang et al, 2005).…”

“…The conventional way of constraining based on interpolation and/or extrapolation can generate unreasonable results in complex geological cases. Wherefore, we can utilize soft constraint based on Markov random field to introduce the prior information (Zhang et al, 2005). The constraint means is still achieved along with geological structure trend.…”

Indicative Factors of Reservoir Sensitivity Based on Pre - stack Inversion

“…Zhang [13] used simulated annealing algorithm of global optimization by combining regularization method and fast simulated annealing algorithm to determine the effects of regularization parameters, and obtain a global optimal solution. Simulated annealing algorithm of global optimization can obtain a global optimal solution, but the calculation is time-consuming.…”