Electrofacies based on conventional logs are found as strongly correlated to core lithofacies, thin section microfacies, and petrophysical measurements (f/k and MICP) as long as the carbonate pore system remains simple. Once dual porosity is present, it is found that neural network using conventional logs can not distinguish 2 rock types having the same range of porosity but different porosity-permeability relation. The dual porosity system is illustrated by strong leaching (i.e. dissolution) overprinting the primary interparticle porosity of a grainstone, and responsible for an increase of one order of magnitude in permeability. The dissolution is observed by patchy feature on core. Similarly, this high level of heterogeneity can be observed downhole by borehole imaging tool. The heterogeneous porosity map from the image tool is then converted into a single curve representing the secondary porosity. This secondary porosity log is added to the conventional logs as input of the neural network model. Then, neural network can discriminate between 2 rock types with same range of porosity but different porosity-permeability relation.
It is nowadays common practice to run conventional logs at all wells, such as gamma ray, density, spontaneous potential, neutron and induction tools, since engineering progresses combined all these probes in one single tool. Cost is negligible compared to valuable information that these logs can bring in term of rock characterization (porosity, mineralogy for instance) and fluid content (gas, water and oil saturation). Statistical methods such as neural network are used to classify the logs responses (i.e. log typing) into a number of electrofacies. If good correspondences exist between on one hand, these electrofacies, and on the other hand, both geological facies and petrophysical properties, then these electrofacies are considered as valid reservoir rock types (RRT). The electrofacies defined from conventional logs at each well are then propagated into a 3D static model, which can be used for dynamic flow simulation.
On the other hand, more advanced tools such as borehole imaging tool or Nuclear Magnetic Resonance (NMR) are run on a limited number of wells for cost reasons. These "unconventional" tools have a much higher spatial resolution: around 6 inches for NMR, and 0.16 inches for borehole imaging (FMI*), opposed to about 2 to 3 feet for conventional tools. It has been shown that further post-processing of FMI image can evaluate physical size of the fracture aperture as small as a few hundred micron1. Beside, Newberry2 introduced a technique called PoroSpect* that allows porosity mapping of the borehole. Primary (e.g. interparticular porosity) and secondary (e.g. dissolution vugs or fractures) porosity could then be distinguished, and this technique has been successfully applied to Gulf Carbonates3. This is of particular interest for carbonate as it is well recognized that unlike conventional siliciclastic reservoir, carbonates have complex pore system distribution. Their complexity lies in the fact that depositional primary pore space is overprinted by both a diagenetical and a stress history, responsible for dissolution, dolomitisation, fractures or stylolites for instance.
However, reservoir engineers do not always see the purpose of integrating such high resolution tools in order to build a static reservoir model whose upscaling will strongly reduce its spatial resolution. Also, the pricing of such high resolution tools limited their use to specific tasks (horizontal wells, cored vertical wells for calibration purpose for instance), while today high oil prices should make their use at entire field scale profitable.