In this paper we use our models to evaluate how the impact of completions varies over the life of a well, across multiple basins. Using an estimate of completions uplift at a single point in time can result in inaccurate forecasts and economics.
|- Tuesday, July 27th at 5:00 PM - Room 371 | Theme 8: Northern Shales - AUTHORS:: T. Cross, D. Niederhut, A. Cui, K. Sathaye, J. Chaplin (Novi Labs)|
Objectives/Scope: Upsized completions have been a primary driver of increased well performance over the last decade. However, it can be difficult to quantify how that impact varies over the life of the well, from peak rate to two year cumulatives or beyond. Understanding that changing impact over days from initial production (IP) is critical to understand well performance and optimize development designs. In this study, we present a machine learning-based analysis of the time-variant impact of completions on oil rates across the Bakken, Eagle Ford, and Wolfcamp unconventional plays.
Methods/Procedures/Process: We trained a series of machine learning models to predict oil production using geology, publicly available completions parameters, and interwell spacing measurements. The model forecasts a vector of production in thirty-day increments over the first three years of a well’s life, enabling us to analyze the impact of the input variables at each timestep. After model creation, we train a surrogate model to produce SHAP Values (SHapley Additive exPlanations), which isolate the contribution of the training features on the model prediction.
Results/Observations/Conclusions: Across each play, the impact of proppant loading decreases through time. In the Bakken, the effect diminishes approximately 50% from the peak at IP 120 through 1080. The Eagle Ford shows a steeper diminishment of impact, falling over 50% from IP 30 through the first year. Proppant has the longest-lived impact in the Wolfcamp, where the rate uplift diminishes by less than 30% over the first three years. Fluid loading similarly declines in impact through time in each basin, though the magnitude of the decay is less. In contrast, well spacing has a larger impact over time, suggesting that operator choices do not always diminish in importance over well life. The larger impact of proppant in early time relative to fluid’s impact may reflect that the primary effect of proppant is to increase fracture conductivity, which will raise early-time production until the limiting factor becomes the flow of fluid to the fracture face. The slower decay of completions impact in the Wolfcamp may reflect increased fracture roughness, which may lengthen the time to fracture closure compared to the other plays. Alternatively, this may reflect a larger stimulated rock volume, which could take longer to drain.
Applications/Significance/Novelty: Understanding how completions impact oil production through time is critical for accurately assessing well economics and for optimizing unit designs. Comparing these time-dependent impacts from basin to basin enables discussions of the geomechanical and reservoir drivers that cause completions impact to last — or to decay rapidly. The approach outlined in this study could be applied to datasets including proprietary completions parameters like cluster count and operational data like artificial lift type in order to derive a more comprehensive understanding of reservoir performance.
Interdisciplinary Components: This work incorporates contributions from data science, reservoir engineering, and completions engineering.