Optimizing estimated ultimate recovery in horizontal wells through data-driven analytics

Abstract

The oil and gas industry has undergone significant changes in completion operation design and execution, largely due to the advancements in machine learning techniques. In our study, we analyzed a dataset of approximately 200 horizontal wells from the Wolfcamp formations, categorized into two groups based on productivity and classification as parent versus child (infill) wells. Using response surface methodology (RSM), a data analytics technique that models the relationship between explanatory and response variables with up to second-order polynomial terms, we illustrate how to optimize horizontal well performance, specifically focusing on Estimated Ultimate Recovery (EUR). We selected a pool of 5 input variables for analysis, including initial oil production of 180 days, injected fluid volume in barrels, number of perforations in each cluster, volume of Hydrochloric Acid injected during frack job in gallons, and volume of injected Linear Gel in gallons.The selections were made in collaboration with industry professionals and validated using data analytics-based methods that rank variable importance. By pinpointing potential strategies to achieve a globally optimal EUR through the adjustment of completion variables (perforations and hydrochloric acid should be set at their data-recorded maximum values of 7.69 and 10.82 on a logarithmic scale), the study aims to simplify the challenging task of maximizing EUR from producing wells, offering an additional avenue for estimating reserves. It emphasizes the adaptability of this approach in allowing different completion designs to be dynamically tailored for optimizing EUR, each with its unique set of inputs. Ultimately, these diverse designs can be synthesized through meta-analysis. In summary, the study provides a practical framework for EUR estimation, readily applicable by operators in unconventional exploration, representing a transformative milestone in the industry. The subsequent phase involves collecting additional data from this specific field area and employing a range of models to improve precision and specificity.