Artificial neural network for permeability damage prediction due to sulfate scaling

Waterflooding is an important oil recovery method, which is used to maintain reservoir pressure and to increase oil productivity. One of the most common problems caused by waterflooding is inorganic scales formation especially barium sulfate scale, which occurs due to incompatibility of injected seawater and formation water, and causes formation permeability decline.Artificial neural networks (ANNs) are new tools, which application of them in petroleum industry has been extended. Since many factors have influence on permeability reduction due to barium sulfate scaling and relation of them with one another is complicated, in this research a model was presented for prediction of permeability damage due to formation of barium sulfate scale using MATLAB software, artificial neural network and waterflooding experimental data in Malaysian and Berea sandstone cores. To design the optimum ANN model, number of neurons, number of hidden layers and training function were studied. Finally, efficiency of the model was evaluated using new data. The proposed artificial neural network predicted permeability and its reduction during water injection with error about 2 percent.

► Using the results of case studies, a feed forward back propagation algorithm was utilized to design the neural networks for prediction of permeability damage due to formation of barium sulfate scale.
► The developed ANN model could predict the permeability with a total average absolute deviation (TAAD) of 0.32% in test stage and 0.91% in performance stage.
► The results of this study showed that the proposed artificial neural network is able to predict permeability decline caused by barium sulfate scale formation with high efficiency.