Like most ultra-tight unconventional gas fields, the shale field appraisal in Sichuan basin in China used production analysis and decline estimates for determining the Expected Ultimate Recovery (EUR) for the initial wells; However, the performance of some wells was significantly different from the behavior of North American plays where the production typically is dominated by square root of time behavior due to the large surface area of the fractures in the multiple-fractured horizontal wells (MFHW). Even the best well in the field, well A, appeared to be controlled by the boundary dominated flow (BDF) when industry accepted rate transient analysis was applied. Although this well was not drilled close to other MFHW, the time to reach the boundaries indicated higher than expected permeability near the well with poorer reservoir beyond these boundaries. Microseismic data indicated that the hydraulic fractures were not strong linear features; rather they appeared to be more complex and distributed in the area typically thought to be the stimulated reservoir volume (SRV). While core data indicated the presence of natural fractures, most were cemented up and initially thought not to be important.
To reduce the uncertainty of the reservoir parameters, an integrated study employing Pressure Transient Analysis (PTA) and Rate Transient Analysis (RTA) was done. The premise of this method is that the time scales for these are not reductant; ergo, they are testing different portions of the reservoir. RTA will use long-term production data, and data are reliable after cleanup and stabilization >1000 hrs. PTA analysis uses shorter term transient data < 1000hrs and is vital to understanding completion parameters. Multiple pressure buildups were done, which not only constrained the fracture conductivity and skin, but also provided insight into how they evolve with production.
A long pressure buildup (PBU) of approximately one month with bottom hole pressure gauges was done to understand the well performance. This data provided insight into why well A did not exhibit the dominant square root of time behavior that everyone expected. The PTA diagnostic plot for the buildup test in shows the transient response was dominated by pseudo-steady state inter-porosity flow (i.e. Warren and Root) from a double porosity model. By matching the buildup test response first with a MFHW including double porosity in the PTA model, the production data could be matched along with the buildup data in an infinite reservoir. The apparent boundary dominated flow is a result from the intrafracture interference through the high permeability portion of the double porosity system. The success from having a coherent model matching both PTA and RTA led to a new approach for estimating a range of EUR values incorporating uncertainties in the reservoir and completion parameters and the performance data. This approach is adapted for well spacing design as well.
The workflow for estimating the range of uncertainty started with a spreadsheet containing the uncertainty on well, completion, and reservoir parameters. For a given set of parameters, the buildup data and the production data were matched with unconstrained parameters and then a production forecast was made for that scenario. For example, one might start the match assuming the fracture height is fully penetrating the reservoir; however, another scenario may assume the heights of the fractures are limited to a smaller fraction of the reservoir. Utilizing this methodology, the smaller fracture height provides a match with longer fracture half-length. The estimated double porosity parameters change as can the interpreted permeability. Other parameters looked at include reservoir properties (e.g. effective permeability, porosity, thickness, kv/kh), fracture half length, effective number of stages, reservoir model types (single layer, multiple layers), permeability degradation, etc. A nice outcome of this process is the comparison of how well each scenario matched the production and pressure transient data. Those that match reasonably well are certainly more plausible than those that cannot match at all.
