Reservoir Porosity Estimation from Well Logs Using Neural Networks

One of the most important factors in reservoir management is the knowledge of sequences and characteristics of the underlying formations. Besides, one of the best methods to reach this goal is well logging, which gives us some important parameters by evaluating formation attributes. To achieve those above mentioned parameters there is the necessity to use other parameters too, but they need much more time and more cost. Thus, proposing a new solution for finding those parameters using just logging data is welcome. In this paper, neural network tool is considered to predict “ porosity” by using DT, LL3, ILD, ILM, RHOB, DRHO, PEF, GG, and PHI. This network is a multilayer perceptron network (MLP) consists of 2 hidden layers including tangent hyperbolic functions as transfer functions. Moreover, output layer consists of one neuron including a linear function as transfer function too. Also, we examine some learning methods for training MLP and the results are reported. We compare our approach by conventional method (core data); the experimental results show that the ANN is a suitable tool for modeling and predicting porosity.

Keywords:

Porosity , Multilayer Perceptron Neural Network (MLP) , Well Log