Carbonate reservoirs usually consist of multiple units each with different degrees of layer-bound fracture density, height, length, and angular scatter. Some units may have finite fracture networks (FFNs), and some others may contain only isolated fractures. FFNs are islands of interconnected fractures within a sea of isolated fractures. The fracture volume can be estimated from the initial surge in production if a wellbore intersects only a single FFN. Combining with fracture density and aperture measurements from borehole images, FFN volume, drainage area, and other relevant attributes can be calculated. If, however, a slanted well intersects an unknown number of FFNs each in a different unit, it is not possible to estimate each FFN fracture volume from initial fracture production. Because the connectivity threshold for initiation of FFNs is not known or cannot be calculated, it is not possible to determine how many FFNs the borehole intersects. One way is to set an arbitrary threshold of fracture connectivity, 𝜆c.
Based on this initial value, a fracture-flow model can be generated and adjusted until the inferred rate time and cumulative time curves converge to observed values. The adjustment is accomplished by changing the number of FFNs using different connectivity threshold 𝜆c and estimating their size by modifying unknown or uncertain parameters such as fracture aperture. The final model yields the mostly likely number and size of all FFNs and provides an approximate forecast for future production performance.
The ability to match production performance without taking pressure into consideration shows that rate decrease in fractured reservoirs with FFNs is not only related to pressure drawdown but also, perhaps more importantly, to depletion of FFNs starting from the smallest size or the one with highest drainage efficiency.
One drawback of the method is that different numbers of FFNs may have similar inferred production performance and cumulative production. This is an inevitable drawback inherent in most models that must rely on multiple uncertain or unknown parameters. The only way to decrease the range of possible solutions is to decrease the uncertainties in the underlying input parameters. In any case, alternative results may be presented as a range of possible solutions.