Porosity Evaluation Using Artificial Neural Network, Optimized with GA and PSO

Porosity in conjunction with some other reservoir parameters plays an integral role in oil and gas reserves evaluation. Porosity is normally obtained through precise laboratory core analysis and empirical correlations from log evaluation. And recently, newer techniques such as CT Scan have been in use to determine such parameters as porosity. However, these techniques are time consuming and costly, therefore those techniques that can extract the measure of petro-physical parameters of the reservoir rock from well logs are of great importance to reservoir engineering studies. Artificial intelligence and machine learning techniques can well proved their place in such functions, hence, well logging data and data mining techniques have been used in this study to determine the influential parameter of porosity. For this, the data have been normalized first and thenmultilayer perceptron neural network (MLP) and radial base function neural network (RBF), both of which are machine learning techniques, have been used to evaluate the vertical and horizontal porosities in an Iranian oil and gas field based on geophysical data. GA and PSO have also been used for optimizing the parameters. Thereafter, the results from these techniques have been compared with each other and discussed. Reservoir data from four different wells in a carbonate reservoir in the south west of Iran have been used. In the input of the artificial neural networks the following log data have been used; SGR, CGR, DT, ILD, Density, NPHI, Caliper in conjunction with Depth, and in the output porosity data from vertical and horizontal cores has been used

Keywords:

Artificial Neural Network , Porosity , Core Analysis , Reservoir