Development of New Permeability Formulation From Well Log Data Using Artificial Intelligence Approaches

2018 ◽  
Vol 140 (7) ◽  
Author(s):  
Tamer Moussa ◽  
Salaheldin Elkatatny ◽  
Mohamed Mahmoud ◽  
Abdulazeez Abdulraheem

Permeability is a key parameter related to any hydrocarbon reservoir characterization. Moreover, many petroleum engineering problems cannot be precisely answered without having accurate permeability value. Core analysis and well test techniques are the conventional methods to determine permeability. These methods are time-consuming and very expensive. Therefore, many researches have been introduced to identify the relationship between core permeability and well log data using artificial neural network (ANN). The objective of this research is to develop a new empirical correlation that can be used to determine the reservoir permeability of oil wells from well log data, namely, deep resistivity (RT), bulk density (RHOB), microspherical focused resistivity (RSFL), neutron porosity (NPHI), and gamma ray (GR). A self-adaptive differential evolution integrated with artificial neural network (SaDE-ANN) approach and evolutionary algorithm-based symbolic regression (EASR) techniques were used to develop the correlations based on 743 actual core permeability measurements and well log data. The obtained results showed that the developed correlations using SaDE-ANN models can be used to predict the reservoir permeability from well log data with a high accuracy (the mean square error (MSE) was 0.0638 and the correlation coefficient (CC) was 0.98). SaDE-ANN approach is more accurate than the EASR. The introduced technique and empirical correlations will assist the petroleum engineers to calculate the reservoir permeability as a function of the well log data. This is the first time to implement and apply SaDE-ANN approaches to estimate reservoir permeability from well log data (RSFL, RT, NPHI, RHOB, and GR). Therefore, it is a step forward to eliminate the required lab measurements for core permeability and discover the capabilities of optimization and artificial intelligence models as well as their application in permeability determination. Outcomes of this study could help petroleum engineers to have better understanding of reservoir performance when lab data are not available.

Author(s):  
Samuel Lucky Arubi ◽  
Bibobra Ikporo ◽  
Sunday Igbani ◽  
Ann Obuebute ◽  
Sylvester Okotie

2020 ◽  
Vol 26 (1) ◽  
Author(s):  
O. Okolo ◽  
B.Y Baha

Selection of a software project is a critical decision. This selection involves prediction to ascertain a project that provides the best business value to the organization. The process of selection is carefully undertaken to optimize scarce resources available, which makes it impossible to simultaneously invest in all business ideas and systems. The current traditional method of software selection does not consider risk factors among the many variables necessary to predict a project that could provide the best business value. More so, the current method such as an artificial intelligence approach, where project managers use more robust models to make predictions have not received the needed attention in developing models for software project selection. This research applied a branch of Artificial Intelligence called Artificial Neural Network to classify projects into three levels. The research designed an artificial neural network of four inputs, one hidden layer with twenty-seven (27) neurons, and three outputs. Keras, a python deep learning library that runs on a theano background was used to implement the model. This research used a secondary dataset, which was enhanced by the synthetic approach, to make the required data features needed in machine learning applications. Backpropagation Algorithm enabled the model to train and learn from the data, and K-fold cross-validation was used to measure the accuracy of the model on unseen data. The results of the simulation showed that the model performed up to 98.67% accuracy with a standard deviation of 2.6% performance on unseen data. The research concludes that using the artificial neural network for software project selection unleashes a new vista of opportunities in artificial i ntelligence where intelligent systems are developed based on robust models from data forproject selection.Keywords: Artificial Neural Network, Project selection, Machine LearningVol. 26, No. 1, June 2019


Sign in / Sign up

Export Citation Format

Share Document