Reference Synthetic Data Set for Drilling Inventory Optimization

Summary Data analytics holds immense importance in optimizing inventory management in the oil field due to its significant business impact. However, the progress of research is limited by restrictions of sharing inventory data publicly, owing to privacy concerns. This paper introduces a novel synthetic data set for drilling inventory systems in oil-integrated companies, which shall offer a valuable open-source reference for future studies. Additionally, diverse analytical approaches were used to showcase their effectiveness in leveraging cost reductions. Generating synthetic data sets involves the implementation of data-driven and process-driven methodologies. Given the inherent risk of endogeneity bias, a process-driven approach, combined with domain knowledge, was followed. Primarily, predefined criteria were established to govern data set generation while accounting for various possible well templates. In essence, multiple constraints outlined the relationship among oil prices, expected annual drilling activities, timing and cost of item receipts, and the interdependence of item quantities and issuances. Next, controlled noise was introduced to account for irregularities and periodic inclusion of cutting-edge items. Ultimately, diverse inventory scenarios were subjected to analytics and optimization. The introduced inventory data set is characterized by randomness, rule-based procedure, and iterative process. It allows replicability and adaptability to divergent layouts by virtue of user-defined parameters. Overall, it possesses a history of transactions that extends over 20 years, exhibits an average asset value of 5 million USD, features around 500 unique stock-keeping units (SKUs), and accommodates more than 15,000 transactions, such as receipts, issues, returns, sales, loans, inspections, and junks. Initial inventory volume was estimated based on the drilling activities of a hypothetical medium-sized oil company that operates an average of 10 drilling and workover wells per year. The generated data set covers abundant specifications of multidisciplinary SKUs, covering tubular goods, drilling accessories, completion equipment, and wellhead materials. Items were selected delicately to represent diverse well configurations, encompassing sidetracks, slim holes, gravel-packed wells, artificially lifted wells, and dual producers. Furthermore, probable mistakes and human errors were taken into account, to improve the system’s reliability to closely mimic real-world conditions. Subsequently, and upon investigating various key performance indicators, it was observed that analytics and clustering techniques are promising in elucidating latent challenges, maximizing profits, and achieving cost savings of up to 10% of the inventory’s total value. In this paper, we present a novel, openly accessible synthetic data set for drilling inventory systems in oil-integrated companies for the first time in literature, to the best of the authors’ knowledge. The introduced data set was meticulously designed to ensure scalability and minimize drift. Positioned as a reference for future research, this data set aims to drive forward the advancement of research in analytics and machine learning, to deliver ingenious solutions for inventory management optimization in digital oil fields.