DQNN: Pore-scale variables-based digital permeability assessment of carbonates using quantum mechanism-based machine-learning

Permeability is a key parameter of rock reservoirs, suggesting the flow characteristics of rock reservoirs. Permeability prediction of carbonate reservoirs is still a great challenge due to its complex pore network and wide range permeability. This work is to establish a digital quantum mechanism-based neural network (DQNN) to study the permeability using the digital porosity, coordination number and pore network size. Experiments and artificial neural network methods (ANN) are applied to validate the accuracy of the proposed DQNN method. In these methods, the pore-scale variables extracted from the micro-CT images of 200 carbonate samples are applied. Results show that the permeabilities obtained from experimental, artificial neural network and DQNN methods agree well with each other. Digital pore size, pore throat size and length are better parameters, while coordination number and porosity are relatively secondary parameters for permeability descriptions of carbonate reservoirs. Compared with the ANN method, the proposed DQNN method is superior in low computation time and high ability for complicated problems.

Automated summary

The research establishes a DQNN to study the permeability of carbonate reservoirs using digital porosity, coordination number and pore network size, with experiments and ANN validating its accuracy. Results show good agreement among different methods in predicting permeability, with digital pore size, pore throat size and length being more crucial parameters.