Well logging curve inversion method

Abstract

The invention relates to a well logging curve inversion method which comprises the steps of (1) selecting a well and a well logging curve of a work area; (2) performing time depth on the well logging curve; (3) extracting a seismic attribute along a well locus; (4) analyzing the interactive relationship of the seismic attribute of the well logging curve; (5) selecting a layer; (6) establishing a stratum grid rack; (7) establishing an initial model; (8) training a nerve network; and (9) outputting a training result. The well logging curve inversion method is advantageous in that geological statistics is well combined with nerve network technology by a well logging curve inversion method according to an RBF radial basis function nerve network algorithm; and multi-attribute-driven crack density inversion based on the geological statistics is realized. The well logging curve inversion method surpasses existing random simulating and multi-attribute well logging curve. The well logging curve inversion method has functions of obtaining a stable inversion result, expressing nonlinear constrain of multiple seismic attributes to the inversion result, and realizing a prediction precision which is higher than that of other software which singly depends on geological statistics.

Description

technical field [0001] The invention belongs to petroleum exploration and development technology, in particular to a well curve inversion method. Background technique [0002] At present, there are two methods of well curve inversion: random inversion of well curve and neural network inversion of well curve. The stochastic inversion method of well curves adopts the co-simulation method, and uses seismic data as "soft" constraints to carry out well curve inversion. The constraining effect of seismic attributes in the inversion depends on the correlation between seismic attributes and well curves. Larger, the greater the confinement effect of the earthquake. The neural network well curve inversion method utilizes the neural network to establish the non-correlation relationship between the well curve and seismic attributes, and uses the correlation relationship to carry out the well curve inversion. The above two inversion methods have certain disadvantages. The random invers...

AI Technical Summary

Problems solved by technology

The above two inversion methods have certain disadvantages. The random inversion method of well curves can only introduce a seismic attribute as a constraint. At the same time, the constraint process is a simple linear mapping. If the correlation between seismic information and well curves is complicated, The constraint effect of seismic information on the inversion results is greatly reduced, and the inversion results are basically interpolation results

The neural network algorithm is not stable enough. If the correlation between the seismic attributes and the well curve is not very obvious, the well curve inversion result of the neural network is often not good, and the multi-solution is very strong

Images