Volume of Clay Estimation Using Artificial Neural Network Case Study: Berkine Basin Southern Algeria

Abstract

The estimation of the volume of clay is one of the most important parameters in the characterization of shaly sand reservoirs, by its impact on the estimation of reserves and the production. The estimation of the volume of clay accurately will allow a better determination of the volume of the matrix which will reduce the uncertainty in the evaluation of the reservoir formation through a better estimation of the formation water saturation and the effective porosity for conventional reservoirs. For this reason, the target of this study is to find an efficient solution of this problem in the reservoir of Berkine basin by the application of artificial intelligence methods on logging data: gamma ray, thorium, uranium, potassium, density, and coupled with the volume of clay measured by X-ray diffraction. Due to the topology (5-11-1) of the model multilayer perceptron neural network adopted to estimate the volume of clay and to validate by numerical performance indices between the simulated values and the observed values (R = 0.99, RMSE = 0.0003, and MAE = 0.0001), it was possible to estimate the missing clay volume of 1390 points which corresponds to the 200 m in the well W_A, and this technique allows the gain of cost and time in the laboratory.