Shale reconstruction method combining convolutional neural network and shale soft data

Abstract

The invention relates to a shale reconstruction method combining a convolutional neural network and shale soft data, and the method comprises the following steps: 1), obtaining a three-dimensional image of a shale sample as a target image, and randomly obtaining the hard data of a part of shale images from the target image; 2) performing interpolation processing on the hard data image to obtain anestimated value of the shale data distribution condition, and taking the estimated value as input soft data; and 3) inputting the soft data image into a convolutional neural network, and outputting areconstructed image. Compared with the prior art, the method has the advantages that the shale hard data is interpolated by using the simple Kriging algorithm to obtain the estimated value of the shale data distribution condition, the estimated value is used as the soft data input by the convolutional neural network, the soft data image is input into the convolutional neural network, the reconstruction result with high accuracy can be obtained, the consumed time is short, and the method is convenient for large-scale application.

Description

technical field [0001] The invention relates to a shale reconstruction method, in particular to a shale reconstruction method combining a convolutional neural network and shale soft data. Background technique [0002] As the storage and flow carrier of shale gas, the rock pore structure of shale reservoirs is complex, and the size of pores varies from nanometers to micrometers, accompanied by naturally developed micro-fractures, and the gas in pores and fractures of different sizes The state of existence is different from the characteristics of movement. The coupling effect of pores, fractures and fluid in shale directly affects the physical and mechanical properties of shale, among which the shale pore-micro-fracture structural characteristics directly affect the energy storage of shale gas reservoirs and the seepage behavior of internal fluids. Due to the small pores and complex mineral composition of shale, accurate and quantitative characterization of shale pore-micro-f...

AI Technical Summary

Problems solved by technology

The above numerical reconstruction method requires a large amount of training data, resulting in heavy CPU load, long running time, and strong randomness of the reconstruction effect.

Images