Prediction of fracture dip using artificial neural networks

Abstract

Fracture characterization and fracture dip prediction can provide the desirable information about the fractured reservoirs. Fractured reservoirs are complicated and recent technology sometimes takes time and cost to provide all the desired information about these types of reservoirs. Core recovery has hardly been well in a highly fractured zone, hence, fracture dip measured from core sample is often not specific. Data prediction technology using Artificial Neural Networks (ANNs) can be very useful in these cases. The data related to undrilled depth can be predicted in order to achieve a better drilling operation, or maybe sometimes a group of data is missed then the missed data can be predicted using the other data. Consequently, this study was conducted to introduce the application of ANNs for fracture dip data prediction in fracture characterization technology. ANNs are among the best available tools to generate linear and nonlinear models and they are computational devices consisting of groups of highly interconnected processing elements called neurons, inspired by the scientists' interpretation of the architecture and functioning of the human brain. A feed forward Back Propagation Neural Network was run to predict the fractures dip angle for the third well using the image logs data of other two wells nearby. The predicted fracture dip data was compared with the fracture dip data from image logs of the third well to verify the usefulness of the ANNs. According to the obtained results, it is concluded that the ANN can be used successfully for modeling fracture dip data of the three studied wells. High correlation coefficients and low prediction errors obtained confirm the good predictive ability of ANN model, which the correlation coefficients of training and test sets for the ANN model were 0.95 and 0.91, respectively. Significantly, a non-linear approach based on ANNs allows to improve the performance of the fracture characterization technology.