Method and application of homogeneous digital core permeability prediction based on TensorFlow

Key words:  digital core  TensorFlow  BP artificial neural network  permeability prediction

Abstract:

The permeability of core samples is usually measured in laboratory using conventional techniques, which is inefficient, tedious and time-consuming. In this study, a permeability prediction method for homogeneous digital cores was proposed based on machine learning. Firstly, a large number of homogeneous digital cores were randomly generated. Their porosity and permeability were calculated by a pore network model, and the results were taken as the sample database for establishing a machine learning model. Then, based on the BP artificial neural network method, the porosity and permeability data of the cores were extracted and analyzed, and used for training the corresponding machine learning model. The accuracy of the machine learning model was verified in comparison with laboratory experiments. The results show that the machine learning model can provide an accurate and efficient method for permeability prediction. The error between the permeability calculated by the model and that measured by experiment is only 3.1%.The machine learning method can be applied in oilfield for core analysis, which can avoid a large number of core testing, and improve the calculation efficiency of core permeability.