Well logging curve continuation method based on deep learning simple circulation unit

Abstract

The invention relates to the technical field of logging engineering, and discloses a logging curve continuation method based on a deep learning simple cycle unit, comprising the following steps: data acquisition: collecting logging data and rock physical property parameters; preparing data: extracting logging data; data of the drilled part of the to-be-continued well is used as a training set, and data of the undrilled part of the to-be-continued well is used as continuation data; building a model: building a neural network model according to the simple circulation unit; continuation of the unknown section logging data: inputting the continuation data into the neural network model, and continuating the unknown section logging data. According to the invention, a simple circulation unit is adopted as a basis for constructing the network, and the module has a simpler structure, fewer parameters and better convergence compared with other circulation neural network modules; and secondly, the continuation method obviously improves the performance of the training network, parallel processing is carried out in the training process, the training speed is greatly improved, and the method also has very high practicability in engineering while drilling.

Description

technical field [0001] The invention relates to the technical field of well logging engineering, in particular to a logging curve extension method based on a simple cycle unit of deep learning. Background technique [0002] In the field of geophysical well logging, well logging curve parameters are one of the basic seismic elastic parameters that characterize reservoirs. The correlation coefficient between well logging curve parameters and reservoir porosity, saturation, and shale content is very high, which can most intuitively reflect the Changes in subsurface reservoirs and fluids. In addition, the log parameter data is related to the calculation of many other important parameters in the geophysical field, such as overburden pressure and so on. Therefore, the accuracy of log curve parameter data restricts the result accuracy of these important parameters in the geophysical field. Therefore, the parameter data of well log curves play a very important role in reservoir de...

AI Technical Summary

Problems solved by technology

[0005] The purpose of the present invention is to provide a logging curve continuation method based on a simple cycle unit of deep learning to solve the problem that the training speed of the existing logging neural network model needs to be improved

Images