Generation method of gan-based well test interpretation proxy model

Abstract

The invention discloses a GAN-based well testing interpretation proxy model generation method, comprising: drawing a well testing curve according to existing pressure data; adjusting the image size of the well testing curve; sampling all pixel points on the image of the well testing curve, Set the pixels less than the given threshold as white, and set the pixels greater than the given threshold as black, so as to highlight the well test curve from the image; set the length of both sides on the image to be 1 unit length, Each pixel is discretized into a coordinate in the [0,1] interval, and the coordinate is recorded as the input data for the next model training. Configure the GAN generative adversarial network; input the recorded data into the model for training, and obtain a well test interpretation proxy model. The advantages of the present invention are: data preparation is provided for the application of the neural network model, so that the small sample data has the feasibility of training the neural network model, and more real proxy model data can be generated.

Description

technical field [0001] The invention relates to the technical field of oil testing and well testing and reservoir interpretation, in particular to a method for generating a well testing interpretation proxy model based on a generative adversarial network (GAN, Generative Adversarial Networks). Background technique [0002] Well test interpretation is a technical means of analyzing production well test data through seepage theory to evaluate formation or production well parameters. It is the practical application of oil and gas seepage theory in oil and gas field development. Because the data used in well test evaluation come from production wells or water injection wells under flowing conditions, the analysis results can better describe the dynamic characteristics of the reservoir. [0003] As early as 1989, Robert Hecht-Nielson proved that any closed interval continuous function can be approximated by a hidden layer BP neural network, which provided the theoretical basis fo...

AI Technical Summary

Problems solved by technology

However, these methods cannot actually deal with well test interpretation image data very well. When the well test interpretation theory processes the image, it focuses on the relative positional relationship between the two curves, and simple translation cannot well The generated new graphics are distinguished from the original data. On the one hand, the establishment of an ideal model is complex and requires a large amount of calculation. On the other hand, the ideal model cannot well explain the actual formation conditions.

Images