Controllable neural network-type method for predicting reservoir permeability

Abstract

The invention provides a controllable neural network-type method for predicting reservoir permeability. The method comprises the following steps: SI, acquiring seismic data; adding fault control by taking known drilling data as a longitudinal control point and seismic sequence interpret horizons as horizontal restraint; establishing a geologic framework model; SII, performing seismic inversion under the control of the geologic framework model through a genetic algorithm to obtain N wave impedance data bodies; SIII, performing phase-controlled cloud simulation reservoir permeability prediction by aiming at each wave impedance data body, wherein each wave impedance data body is used for predicting M permeability data bodies; SIV, sedimentary facies comparison on N*M permeability data bodies and mutual linear analysis among the data bodies; SV, performing equal-weighted treatment on the N*M permeability data bodies to obtain final permeability data bodies; SVI, performing engraving aiming at an oil reservoir part on the final permeability data bodies to provide data for a downstream reach.

Description

technical field [0001] The invention relates to a geophysical prospecting reservoir prediction method, in particular to a method for predicting reservoir permeability using effective seismic data information, especially a controllable neural network method for predicting reservoir permeability. Background technique [0002] Oilfield development has entered the middle and late stages, and the accuracy of digital modeling plays an increasingly important role in the design of development well patterns and the maximization of injection and production benefits. How to accurately predict reservoir permeability has always been the pursuit goal of reservoir geophysicists. At present, the method of reservoir permeability prediction is mainly based on the inversion of seismic data and the processing of some amplitude information. There are two main types of reservoir permeability prediction, one is geological statistical simulation, and the other is by means of inversion Perform line...

AI Technical Summary

Problems solved by technology

These two types of reservoir physical property prediction methods have their own advantages and disadvantages. The geostatistical simulation prediction results make full use of logging information, but there are many uncontrollable factors in the no-well area; although the method of linear conversion of inversion data uses Seismic data information has certain constraints in areas without wells, but the disadvantage is that the prediction accuracy is low

Another method is to directly use seismic information and logging information (such as permeability data) to establish a well-seismic relationship, and predict permeability with the help of seismic data. However, it is usually difficult to find a matching linear relationship between seismic traces and permeability curves, and the predicted results are very poor. Difficult to achieve satisfactory results

Images