Abstract
The topic of fracture complexity is commonly evoked when discussing hydraulic fracturing of unconventional reservoirs. In this context, it is typically considered beneficial to successful stimulation, as it provides increased surface area, relative to single planar fractures. However, in the near-wellbore region (NWR), this same fracture complexity, commonly referred to as tortuosity, can be detrimental to successful placement of fluid and proppant. In the extreme, if not properly identified and mitigated, fracturing stages may need to be abandoned which leads to unstimulated sections of the wellbore and reduced completions efficiency. Yet, the ability to adequately quantify this phenomenon during stimulation remains limited.
In this paper, we show how modern diagnostic techniques can be leveraged to provide insight into this critical region. Specifically, we combine interpretations from both fiber optic distributed acoustic sensing (DAS) and external downhole pressure gauges (BHG) to improve the characterization of the NWR. This project was executed during the stimulation of a horizontal well located in the Wolfcamp formation within the Midland Basin. We first review observations from the external cemented in place fiber and the external pressure gauges.
The second section presents an investigation of fracturing net pressures trends identified with external pressure gauges. We apply traditional Nolte-Smith fracture diagnostics to analyze fracture propagation and near-wellbore proppant dynamics. The net-pressure investigation reveals that even in unconventional reservoirs, Nolte-Smith diagnostic plots are applicable, when external pressure gauges are available. We show that near-wellbore proppant screen-outs identified by the Nolte-Smith plot are independently identified by Distributed Acoustic Sensing (DAS) data.
In the third section we have attempted to develop a process to quantify near-wellbore tortuosity, where machine learning (ML) algorithm(s) were utilized to estimate the friction pressure induced by near-wellbore tortuosity. The training, testing and validation needed for machine learning algorithm(s) were based on utilizing DAS data, downhole gauge data, pumping schedule and post fracturing reports. The studies indicate that friction pressure due to tortuosity is initially high within the transient rate period and decreases to stable values later within the stage. The validation studies show promising performance of ML algorithm(s) for near-wellbore friction pressure estimation, even without downhole gauge data as inputs. It is expected that with further development of ML algorithms needing limited training data shall allow development of diagnostic tools for better prediction of bottom hole treating pressures in wells without the need of acquiring high frequency downhole data.
The paper also makes an attempt to validate the application of Nolte-Smith plot in unconventional reservoirs, especially in characterizing the NWR. Additionally, fluid communication between stages highlights the importance of the NWR on ensuring stage isolation. Finally, the applied ML algorithm for near-wellbore tortuosity pressure estimation is shown to have a reasonable generalization performance, which may serve as a diagnostic tool for completion optimization.
