Residual oil distribution prediction method and device based on deep convolutional neural network

Abstract

The invention provides a residual oil distribution prediction method and device based on a deep convolutional neural network, and belongs to the technical field of oil development. The prediction method comprises the steps of simulating and establishing a training data set according to geological parameters and development parameters of an oil reservoir of a type to which a target oil reservoir belongs; training by using the training data set to obtain a deep full convolutional coding and decoding neural network prediction model; and predicting the remaining oil distribution of the target oil reservoir at any moment by adopting the deep full convolutional coding and decoding neural network prediction model. According to the prediction method, through geological parameters and development parameters of a target oil reservoir, factors influencing remaining oil distribution and influences of time sequences are considered, different numerical simulation schemes are set for the oil reservoir of the type, and a data set is established through oil reservoir numerical simulation; and then a deep full convolutional coding and decoding neural network is used as a framework for training, so that rapid and accurate prediction of remaining oil distribution of the oil reservoir at any moment is realized, and guidance is provided for efficient development of the oil reservoir.

Description

technical field [0001] The invention relates to the technical field of petroleum development, in particular to a method for predicting remaining oil distribution based on a deep convolutional neural network and a device for predicting remaining oil distribution based on a deep convolutional neural network. Background technique [0002] Water injection development is the most widely used secondary oil recovery technology. It injects water into the reservoir to increase the energy of the reservoir, thereby improving the development effect. However, after decades of development of conventional oil reservoirs in China, most oil fields have entered a period of high water cut, with an average water cut of over 90%, resulting in reduced water flooding efficiency, serious low-efficiency water circulation, and poor water injection development results. At the same time, the distribution of remaining oil in the reservoir is very complicated, showing the characteristics of "overall disp...

AI Technical Summary

Problems solved by technology

This method only considers the influence of the change of injection-production system on the distribution of remaining oil, and cannot predict the distribution of remaining oil saturation under different reservoir parameters (such as permeability distribution, initial oil saturation distribution, etc.)

[0007] At the same time, the above methods based on artificial intelligence to predict the distribution of remaining oil saturation can only predict the distribution of remaining oil for a specific reservoir, while for other reservoirs with similar geological characteristics and reservoir parameters, it is necessary to re-establish the model Training, resulting in low prediction efficiency of the method, poor model versatility, and does not reduce the amount of calculation when applied to other reservoirs

[0008] In addition, existing methods can only predict the distribution of remaining oil saturation at the time points included in the sample set, but cannot predict the distribution of remaining oil saturation beyond these time points

Images