A significant improvement in the prediction of permeability from wireline logs has been achieved for the Niger Delta reservoirs using the concept of Flow Zone Indicators (developed by Amaefule et al) based on genetic unit classification as well as the application of neural networks. This concept is building on earlier reservoir description work whereby reservoirs in the Niger Delta have been classified according to the environment of deposition and by the lithofacies associations (i.e. genetic units) in these depositional environments. Using core data, averages of the FZIs have been computed for each genetic unit. FZI values have then been assigned to reservoir intervals without core data whose respective genetic units have been identified from log data using for example neural networks. The permeability values in the uncored reservoir intervals have thus been estimated using the permeability-FZI-porosity relationship for identified genetic unit in those wells or through multiple nonlinear regression using neural networks. Clear benefits of this improved estimation of permeability have been achieved in the process of history matching well behaviour in full field reservoir simulation models.
A successful reservoir model requires that the reservoir be properly characterised in terms of the lithology, porosity, permeability and hydrocarbon saturation. Knowledge of the permeability distribution is critical to field development planning. Recent experience in SPDC has shown that there is a disparity between the calculated permeability and that measured on cores. Up to 1996 the most widely applied technique for permeability modelling in SPDC was a modified form of the Kozeny-Carman equation (Ref.: 1). This model has however been observed to underestimate permeability in the very good quality sands and to overestimate the permeability in the low quality sands. This observation was recorded lately for the cored intervals of several wells one of which is highlighted in Figure 1.