Smart Hydraulic Fracturing: Optimizing Production from Tight Reservoirs

Abstract An Operator in Middle East conducts 50+ hydraulic fracturing operations annually in tight reservoirs. Despite the revolutionary impact of hydraulic fracturing, traditional optimization methods are computationally intensive and fail to capture real-world complexities, resulting in production losses. To address this problem, we have developed Smart Hydraulic Fracturing (SHF) optimization, a machine learning solution. SHF forecasts short-term production, and optimizes fracturing and completion parameters to maximize production, addressing operational inefficiencies in hydraulic fracturing. The solution ingests a large amount of data related to well attributes, well logs, production, fracture design, and completion parameters. The model is trained to forecast short-term production rates or cumulative values for each hydraulic fracturing stage for up to 2-year period. Furthermore, advanced algorithms optimize hydraulic fracture placement and completion parameters towards achieving maximum / minimum objective function for a given well data with multiple stages and reservoir properties. This approach ensures a robust prediction and optimization workflow, streamlining the enhancement of hydraulic fracturing operations without the need for time-consuming, tedious numerical and analytical methods that often involve high uncertainties and lead to unreliable results. The solution was piloted in one of the operators reservoir involving over 25 wells and 200+ fracturing stages combined. Input data underwent preprocessing, and production was allocated to each stage using data from historical production logging. Subsequently, the model underwent training, testing and validation. The results demonstrated the model's ability to reliably predict production profiles, validated against unseen wells. Furthermore, the fracture design and completion optimization results generated by the solution were reviewed and confirmed by subject matter experts, aligning with historical reservoir and fracturing / completion design experience. The next phase involves scaling the solution across the entire field, training it on a larger number of wells and putting the solution on production to be used by the operator's hydraulic fracturing planning and optimization team. This paper introduces SHF, a novel machine learning solution tailored for operator's hydraulic fracturing operations in tight reservoirs. The SHF, upon full-scale deployment, will enable the operator to rapidly, accurately, and effectively design and optimize hydraulic fracturing parameters within minutes, a massive improvement over the conventional method that takes days. Most importantly, it is projected to maximize recovery in challenging tight reservoirs and cutting operational costs through consistently executing optimal hydraulic fracturing operations.