The Prediction of the Permeability Ratio Using Neural Networks

Abstract

In this research, two novel models were presented for prediction of permeability ratio and its reduction due to the formation of calcium and strontium sulfate scales during waterflooding using multilayer perceptron, radial basis function network, and coreflooding test data. To achieve the maximum efficiency, number of neurons, training function, and activation function were optimized for a multilayer perceptron model and spread parameter was optimized for a radial basis function model. The radial basis function model only could predict trend of permeability ratio reduction in the performance stage, but the multilayer perceptron model predicted permeability ratio after waterflooding with a total average absolute deviation of 0.56%.