Estimation of permeability from a hydrocarbon reservoir located in southwestern Iran using well-logging data and a new intelligent combined method

The most important tools of reservoir management are "description of reservoir properties " and "reservoir simulation ", among which permeability is the most important factor for accurate description and modeling of the reservoir. Standard method for determining permeability in oil industry is core analysis and well testing. These methods are very expensive and not all wells in a field have cores. As a result, the methods applicable to present the petrophysical properties of reservoir, including permeability and well-logging, are highly important because well logs are usually available for all wells in a field. Artificial intelligence methods are new, low-cost and accurate methods that can indirectly estimate the permeability of reservoir in the shortest possible time using well-logging data. Therefore, in this study, using different well logs and a new intelligent combined method of relevance vector regression with Grey wolf optimization (RVR-GWO) algorithm, the permeability of a hydrocarbon reservoir in southwestern Iran (Azadegan oil field) was indirectly estimated. Then, the performance of this hybrid model was compared with that of relevance vector regression (RVR) method. The database consisted of 2506 well-logging data, which were divided into two categories of training data (1754 data) and test data to evaluate the models (752 data). The results showed very good performance of the combined method of RVR-GWO algorithm in estimating permeability. Therefore, this model can be used as a powerful, fast, and accurate method for indirect estimation of permeability in reservoirs where permeability through the core is not measured.