Reservoir Characterization Using Local Linear Models (LLMs) Approach Applied on Seismic Attributes

In this research we used one of the most advanced neuro-fuzzy algorithms called Local Linear Models (LLMs) to estimate the resistivity log from seismic attributes in an oil field in south-west of Iran. Local Linear models belong to the class of intelligent Neuro-Fuzzy approaches which are based on a neuron growing strategy. This method has been developed primarily for the purpose of control in electronic engineering, as it has been shown here, it also can be used as a strong predictor of estimating reservoir characteristics by its application on seismic attributes. In addition, this method can act as an appropriate substitute for conventional methods that have been applied in petroleum industry previously. The input space dividing strategy, used in this method, makes it less affected by the choice of seismic attributes for multi-attribute analysis. Also, this approach can be applied when we have limited number of wells.