Remaining oil distribution prediction method based on autoregressive network model

Abstract

The invention discloses a residual oil distribution prediction method based on an autoregressive network model, which belongs to the technical field of oil reservoir development and comprises the following steps: analyzing main influence factors of residual oil distribution from a basic seepage differential equation of fluid flow; constructing a sample library by using a numerical simulator; constructing an autoregressive network model of a convolutional neural network and a convolutional long-short-term memory kernel, and capturing a complex nonlinear mapping relation between input data and output data; training the constructed neural network model in a training set; evaluating the performance of the trained proxy model by using the minimum absolute value error L1 and the relative L1 error in the test sample set; and outputting an autoregression network model which is trained completely and has good evaluation performance, collecting oil reservoir data in real time, inputting the model, and predicting remaining oil distribution in real time. The remaining oil distribution prediction time can be greatly shortened, and then the time of the oil reservoir automatic history fitting process needing multiple oil reservoir production prediction is shortened.

Description

technical field [0001] The invention belongs to the technical field of oil reservoir development, and in particular relates to a method for predicting remaining oil distribution based on an autoregressive network model. Background technique [0002] When applying the numerical simulation method to calculate the reservoir performance, due to the limitations of people's understanding of the geological conditions of the reservoir, the physical parameters of the reservoir used in the simulation calculation may not be able to accurately reflect the actual situation of the reservoir. Therefore, There will still be some differences between the simulated calculation results and the actual observed reservoir dynamics, sometimes even greatly. The dynamic forecast made on this basis is bound to be incompletely accurate, and may even lead to wrong conclusions. To reduce this discrepancy and make performance predictions as close as possible to reality, history-fitting methods need to be...

AI Technical Summary

Problems solved by technology

[0006] The existing reservoir remaining oil distribution prediction proxy model can only consider the reservoir geological static parameters, and the reservoir remaining oil distribution prediction proxy model cannot be used for the dynamic parameters, and the traditional reservoir numerical simulation calculation involves many grids and calculation Due to the shortcomings...

Images