Diagnostic fracture injection testing (DFIT) is an invaluable tool for evaluating reservoir properties in unconventional formations. The test comprises injection of water over a very short time period, initiating a fracture at the end of a well's horizontal section, followed by a long shut-in period. Analysis of the falloff data with the G-function plot reveals the fracture closure pressure, and the fracture pseudolinear-flow period leads to the initial reservoir pressure.
In most tests, wellhead pressure (WHP) measurements are used because of cost considerations. A wellbore heat transfer model is used to allow conversion of WHP to bottomhole pressure (BHP) by accounting for changing fluid density and compressibility along the wellbore. This model, in turn, allowed us to assess the quality of solutions generated with the WHP data. For DFIT analysis, we adapted the modified-Hall plot for the injection period, whereas both the pressure-derivative and G-function plots were used for the analysis of falloff data. The derivative signature of the modified-Hall plot allows unambiguous estimation of the fracture breakdown pressure (pfb) during the injection period. As expected, the pfb always turns out to be higher than the fracture closure pressure (pfc), estimated with the two methods during pressure falloff, thereby instilling confidence in the solutions obtained.
A statistical design of experiments with coupled geomechanical/fluid-flow simulation capabilities showed that the formation permeability is by far the most important variable controlling the fracture closure time. Mechanical rock properties, such as Young's modulus of elasticity and the Poisson's ratio, play minor roles. In microdarcy formations, a longitudinal fracture takes much longer to close than its transverse counterpart.
Many equiprobable solutions exist while history matching a reservoir's performance, given the ill-posed nature of the inverse problem. To mitigate some of the uncertainty issues stemming from the initial static reservoir description, this study shows how continuous learning evolves when a slate of analytical tools are used while interpreting real-time surveillance data. The combined approach involving the use of analytical tools in conjunction with numerical simulations helps understanding reservoir performance, which, in turn, allows insights into history matching. Specifically, we demonstrate the use of various analytical tools to learn about (a) time-dependent behavior of both producers and injectors with rate-transient analysis to assess an evolving waterflood, (b) reservoir heterogeneity with pressure-transient analysis, (c) degrees of time-variant injection support with the reciprocal-productivity index, (d) injector-producer connectivity with the capacitance-resistance model, and (e) real-time injection-well behavior with the modified-Hall analysis.
The benefits of collective use of analytic tools demonstrate that they should be used either simultaneously or preferably before undertaking a detailed numeric flow-simulation study, particularly where real-time data are being gathered. In particular, the lack of performance match for the entire history with a numerical model becomes transparent when the learning from analytical tools is juxtaposed. This understanding paves the way for much improved learning of reservoir plumbing.