Article Preview
TopMotivation
Placement of transverse HF stages along horizontal wellbores in unconventional gas reservoirs has demonstrated to be an effective approach to extraction of commercial quantities of natural gas from ultra-tight shale formations (Holditch, 2007). Nonetheless, depending on the reservoir’s petrophysical properties, the optimal number of wells as well as optimal spacing, half-length, and number of HF stages may vary considerably. Various development strategies might yield quite different economic results. Moreover, sub-optimal placement of several horizontal wells next to each other might cause them to interfere and produce lower than commercial quantities of gas. Therefore, simultaneous placement of several horizontal wells and optimization of HF stage parameters (including the number, spacing, and half-length) are instrumental for higher profits.
This article presents the integrated optimization framework that has no equivalents in unconventional gas engineering and allows to place wellbores and HF stages along them in a systematic manner. The novel framework uses the discounted NPV objective function as the major comparative base for all production configurations. Among specific objectives of this article are:
- 1.
To discuss the assumptions and applicability of the shale simulation model;
- 2.
To outline benefits of the optimization framework and its limitations;
- 3.
To develop and test the GA workflow that places HF stages along a single well and then extend it to fully integrated framework that distributes traditional zipper-fracs along multiple wellbores (Jacobs, 2014).
TopEvolutionary optimization and GA specifically have been applied routinely in essential petroleum reservoir management problems such as vertical and horizontal well placement. Choosing locations for vertical oil/gas production wells and water/gas injection wells can be solved with gradient-based optimization methods, but such algorithms do not always yield meaningful results and often lead to local optimal solutions (Bittencourt & Horne, 1997; Montes et al., 2001). The problem exacerbates even further in presence of geological uncertainty and risk attitude of the oil producer (Guyagular & Horne, 2001). In these cases, the authors can deploy stochastic evolutionary gradient-free optimization algorithms that start by random sampling of a large parameter domain. These optimization methods show better computational performance, achieve higher values of the objective function, and provide meaningful results to integer and mixed programming problems (Forouzanfar et al., 2010).