Abstract

This research presents a methodology that enables completion engineers to improve hydraulic fracturing treatments by leveraging machine learning alongside visualization of acoustically measured data to pinpoint diminishing marginal returns. This method employs hydraulic horsepower-hour per barrel of slurry volume as a key input estimation metric and integrates a new approach to calculating fracture contact area combined with an acoustically derived measure of fracture connectivity. Such integration facilitates comprehensive trade-off analysis in hydraulic fracture design. This study analyzed 31 wells in the middle Bakken, separated by simulfrac and zipper frac completion types. For simulfracs, data suggest that 5-7 bpm/cluster with an input of 2-3 hydraulic horsepower-hour per barrel of slurry volume yields optimum near field connectivity and fracture contact area. Data for zipper fracs indicates a tradeoff between generating near field connectivity and fracture contact area with a larger range of optimum rate/cluster and hydraulic horsepower-hour per barrel of slurry volume.

Machine learning models demonstrate that the application of 5.5-inch lateral liners can lead to a significant reduction in the average surface treating pressure by approximately ∼500 psi per stage. This decrease translates into a conservation of approximately 74 gallons of diesel and a reduction of 0.75 metric tons of CO2 emissions per hour of pumping, thereby mitigating environmental impacts. The equates to almost 18 metric tons of CO2 emissions per 24-hour period of pumping. Although it's obvious increasing the casing size will decrease treating pressure, there are numerous other factors that will influence pressure. Multivariate machine learning models control for these other factors and the benefit comes from knowing how much pressure will decrease to properly assess the tradeoffs. So, the expected savings in fuel and emission should be considered with the tradeoff of increased casing costs. The study synergizes acoustically sourced data, machine learning predictions, and economic assessments to enhance hydraulic fracturing methodologies.

Introduction

Historically, the optimization of completions designs has been through trial and error (Benish, et al., 2024). The general progression of completion design started with fluid and proppant design, progressed to stage design, and has ultimately led to specific cluster and perforation designs (Lorehn III, Cooper, Nisamidin, & Chalmers, 2024). This progression has evolved over the business and price cycles where operators are now focused on decreasing costs while maintaining the status quo for perforation and cluster efficiency which has been aided by advances in technology. For example, Lorehn III et al. (2024) used acoustic-based imaging technology to determine the effects of different perforation and cluster designs to identify the best design to maintain perforation efficiency (Lorehn III, Cooper, Nisamidin, & Chalmers, 2024). Althani and Lange (2023) identified progression in treatment efficiency as a driver of increasing production with relatively constant proppant per horizontal foot (Althani & Lange, 2023). Given this progression towards a concentration on costs, the industry is now looking at the margins to improve designs and operations. For instance, Fulks et al. (2023) implemented start/stop technology on frac fleets (similar to start/stop technology in vehicles) to reduce fuel consumption and emissions from idle time during operations (Fulks, Shelton, & Bishop, 2023). This shows that although industry-shaping advancements in drilling and hydraulic fracturing have allowed for economic extraction of oil and gas from source rocks, there is still opportunity for further value extraction (Lorehn III, Cooper, Nisamidin, & Chalmers, 2024).

This content is only available via PDF.
You can access this article if you purchase or spend a download.