Forecast Of Oil-water Relative Permeability And Capillary Pressure Curve Of Integral Oilfield Based On Data Driving

Capillary pressure and oil-water relative permeability curve are important data for reservoir numerical simulation and oilfield development prediction.At present,capillary pressure and oil-water relative permeability curve are usually measured by indoor physical experiment methods,which have the problems of long time consuming and high cost.This paper will characterize capillary pressure and oil-water relative permeability curve,which are from integral oilfield,and establish the prediction models by machine learning method,to realize the application of data mining technology in the field of oil and gas field development.The main contents of this paper are as follows:(1)characterization of capillary pressure and oil-water relative permeability curve.Firstly,capillary pressure and oil-water relative permeability curve are normalized by the characteristic points extracted from the capillary pressure and oil-water relative permeability curves.Then,the curves are characterized by linear chain code.Next,the variational autoencoding models are established to compress the curve codes.Finally,the characteristic points and compression values of the curve are extracted as the characterization values of the curve.(2)The establishment of sample set.Under the guidance of oil reservoir theory,the macro geological parameters,fluid characteristic parameters and experimental condition parameters are cleaned and fused to form a learning sample set of capillary force and oil-water permeability curve.(3)The establishment of the prediction models of capillary force and oil-water relative permeability curve.Based on the artificial neural network algorithm,the prediction models of capillary pressure and oil-water relative permeability curve are established,and the sample set data is used for training to optimize the model parameters.(4)Prediction of capillary pressure and oil-water relative permeability curve.The curve feature point values and code values are predicted by the capillary force and oil-water permeability curve prediction model.Next,the coded values are decoded and un-normalized to get prediction curves,which realize the prediction of the capillary force and the oil-water relative permeability curve.In addition,a stepped prediction mechanism is established for the oil-water relative permeability curve prediction lacking capillary force curve data.(5)The program for predicting capillary pressure and oil-water relative permeability curve is compiled.According to a series of data collection and processing,as well as the establishment and prediction process of capillary pressure and oil-water relative permeability curve prediction model,the prediction model of capillary pressure and oil-water relative permeability curve is compiled by python to realize the automation of capillary pressure and oil-water relative permeability curve prediction.The test samples are used to test the prediction ability of capillary force and oil-water relative permeability curve prediction model.The result shows that the prediction error of capillary pressure curve prediction model is 5.27%,the prediction error of oil-water relative permeability curve prediction model is 6.19%,and the prediction error of oil-water relative permeability curve step prediction mechanism is 9.31%.It shows that the prediction effect of capillary force and oil-water relative permeability curve is good.The prediction software of capillary pressure and oil-water relative permeability curve has perfect function and normal operation,which can be used in practical application.