Publications

This paper presents a permeability calculation by integrating geological and reservoir engineering considerations into a supervised learning method. The authors previously calculated permeability log curves for a sandstone reservoir of Abadi field by using parametric equations, and realized that establishment of the universal model was an illusion. As an alternative solution, the authors adopted k-nearest neighbor (kNN) method that is based on a proximity search in the variables space of the training data. The candidate curves, which were individually produced by facies groups, were then mixed together while minimizing of the deviation between permeability - thickness products (kh) by DSTs and same by logs. The resultant log provided better prediction and minimal permeability multipliers. The multi-disciplinal collaboration enhanced the robustness of the method.