Hydrocarbon reservoir parameter estimation using a fuzzy Gaussian based SVR method

Measuring effective porosity helps in evaluating the capacity of rock to contain fluid. In this paper, we estimated porosity in one of the oil fields in southern Iran. For this purpose, we used a well-known method of Support Vector Machine due to its ability to produce models with less risk of overfitting and a good generalisation capacity. Considering that used data are always contaminated with noise, we combined this method with a fuzzy system using membership functions. Different types of membership functions are available and they are chosen on the basis of data distribution. Membership functions add importance and prioritise data points. Each data point is evaluated with respect to the whole data and receives a coefficient between 0 and 1. Data points with coefficient closer to zero have a lower priority in the algorithm and data points with coefficient closer to one have a higher priority and are more important in the algorithm. To compare the fuzzy system effect, the coefficient of determination is calculated for the model including noise. A lower priority is attributed to random noise in data with respect to normal data. The result shows that using fuzzy systems notably improves the robustness of a model in the presence of noise.