Integration Between Different Hydraulic Fracturing Techniques and Machine Learning in Optimizing and Evaluating Hydraulic Fracturing Treatment

Abstract

ABSTRACT Fracture geometry and conductivity are critical parameters for fracture treatment optimization, especially in cases that close to unwanted zones either water-bearing or gas zones. This study investigates the Artificial Neural Network (ANN) model for hydraulic fracturing optimization. The workflow begins with an integrated ANN model, then sets of variable fracture parameters and formation rock properties were utilized for training and testing the ANN based on the most appropriate activation function, the number of hidden layers and the number of neurons. The ANN model considers a 59 real field data of hydraulic fracturing treatments across the western desert of Egypt. The proposed ANN trained based on pressure transient test analysis that was conducted on the real field data. The available data was divided as 70% for training, 15% for validation, and 15% for testing. The optimum number of hidden layers and neurons was achieved after several trials. The proposed ANN model result was promising as compared with the common fracture simulation software. The cross plot of the actual fracture geometry parameters versus the predicted ANN outputs showed a good match with the correlation coefficient (R) for the whole data is 0.93. Then the relative importance of the ANN input parameter on the fracture geometry optimization was employed by the Garson method. The result of this work shows the potential of the approach developed based on the ANN model for predicting the fracture geometry.