APA Corporation is the holding company for Apache Corporation, an American company engaged in hydrocarbon exploration. It is organized in Delaware and headquartered in Houston. The company is ranked 595th on the Fortune 500.
The Wadi Rayan oil field has experienced inconsistent production. This stems from lateral and vertical heterogeneities within the sandstone reservoirs, which profoundly impact hydrocarbon flow. A comprehensive multiscale study integrating sequence stratigraphy, seismic interpretation, and petrographic analysis was conducted to characterise the reservoir system. Our results reveal a complex, tidally-dominated estuarine system deposited during the Cenomanian on a low-gradient shelf. Crucially, reservoir quality is facies-controlled, with high-energy channel sands preserving good porosity (0.15-0.28) despite silica cementation, whereas moderate-energy tidal bars suffer significant porosity reduction from calcite cementation. Structurally, the main field closure is a NE-asymmetrical anticline bounded by faults. We constructed a 3D static model using detailed facies and property distributions, estimating cumulative stock tank original oil in place of 16.33 MMSTB across three reservoir intervals. The model definitively identified a critical stratigraphic barrier-composed of low-permeability shelf shale and carbonate facies-that compartmentalises the reservoir and explains the presence of dry wells (e.g., WR-2X and WR S-1X) at structural highs. This finding challenges the notion that structure alone controls entrapment, demonstrating instead that stratigraphic complexity can override the structural template. We consequently propose drilling two appraisal wells to further evaluate future planning. This integrated approach conclusively links dry wells to specific stratigraphic complexities, enhancing reservoir understanding and providing a predictive framework for improving recovery in similar geological settings.
While conventional bit balling is readily identifiable, the deep formations of Egypt's Western Desert present a unique and subtle variation: unconventional bit balling occurring at depths up to 15,000 ft in high-frequency interbedded lithologies. This paper details the culmination of a five-year investigation challenging prevailing misconceptions regarding bit balling in High-Performance Water-Based Mud (HPWBM) applications, where standard indicators—such as a completely balled bit face—are often absent. We present the deployment of an AI system using a pioneering edge-deployed machine learning model capable of early detection of balling events. By analyzing micro-trend fluctuations, the system isolates the balling signals from lithological noise and ROP control effects. Upon identification, mitigation involves pumping specialized anti-balling pills and employing specific on- and off-bottom drilling practices to clean the bit in situ, thereby reducing invisible lost time (ILT) and avoiding unnecessary bit trips. Furthermore, the integration of real-time monitoring with historical offset data marked a step-change from reactive to proactive optimization. This enabled the operator to pinpoint exact dysfunction depths in offset wells and pump precautionary pills ahead of these zones. The result is a proven capability to predict the precise initiation of balling, enabling intervention before premature bit failure occurs. This strategy not only decreased balling frequency but also enhanced ROP and eliminated time-consuming, after-the-fact recovery operations. The method was successfully implemented across the operator's assets, yielding record-breaking drilling runs in several fields.
Abstract Despite significant progress made in drilling automation over the last decade, autonomous drilling of a hole section remains at an early stage. Autonomous drilling cannot happen without near 100% situational awareness (SA). Going by the adage “what cannot be measured cannot be improved”, the first step in the process towards autonomous drilling is measuring the SA capability of drilling advisory models. The objective of this paper is to demonstrate SA measurement with example use cases. We utilize the Situation Present Assessment Method (SPAM) to evaluate the SA capability of a drilling advisory model on historical datasets from land operations in the Middle East. SPAM measures SA by querying the model outputs in real-time during the historical playback to assess awareness under operational conditions. Responses and response times are used to evaluate the three levels of SA (perception, comprehension and projection). We also apply a Large Language Model (LLM) to evaluate the SA of the model in a dynamic, interactive scenario, such as a simulated conversation with the end user. The questions for testing the model were carefully crafted to test all 3 levels of SA. An example of a level 1 SA question that aims to test perception is “What is the current status of the sensors used?” An example of a level 2 SA question which tests comprehension is “Why is the pipe stuck? A level 3 SA question on the other hand tests the projection capability. They are usually of the type “What should be done next”? On the datasets used in this study, the questions were posed roughly after every 30 minutes of historical playback in drawing statistics and test results. The use of LLM allowed for the questions to be refined. After each such query a subject matter expert evaluated the answers provided for all the questions. The model used in this test achieved varying SA accuracies for the three levels. As expected, when additional sensor data was made available in a particular test run, the SA accuracy increased. Also, level 3 SA questions were better answered when offset well data was available. This paper outlines an approach to measure the SA capability of event detection (drilling advisory) models, which is key to making progress towards autonomous drilling. The approach is integrated with an LLM to facilitate testing the SA capability in a conversational manner. This is the first paper that presents an approach to quantifying the SA capability of a drilling advisory model.
High impedance (hard) mudstones are sometimes observed in association with sand injection complexes in the Paleogene petroleum province of the northern North Sea. A hard mudstone surrounding a water-bearing sandstone can give a similar acoustic response to an oil-bearing sandstone surrounded by low impedance (soft) mudstone. The presence of hard mudstone thus impacts the ability to predict hydrocarbon presence directly from seismic data during exploration. To establish the mechanism of 'hardening' to better predict the presence of variable mudstone characteristics, we examine three cored wells from the Beryl Embayment. Well logs and core were examined to characterise the structure, petrology, petrophysical properties and spatial distribution of both hard and soft mudstones. The results indicate that mudstone hardening is most likely associated with mechanical compaction and efficient dewatering of mudstones into the sand injection complex. This process is enhanced where sand injection complexes transect primary overpressure zones, that promote dewatering from basal overpressured mudstone into the injection network. This study highlights that seismic response needs careful investigation in the context of the complexity of the injectite complex along with variable mudstone attributes. Additionally, this process highlights the role sand injection complexes play in efficient dewatering through lateral transfer in overpressured basins.
Plug and perf hydraulic fracturing is performed with high-pressure injection of fluid and proppant from perforation clusters along a wellbore. During this process, uniform placement of fluid and proppant is important for maximizing economic performance. In prior work, we developed a wellbore-proppant transport simulator, which accounts for a wide range of phenomena, including proppant suspension, proppant settling, perforation erosion, perforation pressure drop, inertial effects, perforation orientation, and random variance, among others. In the present work, we calibrate the simulator to downhole imaging measurements of perforation erosion from wells in the Midland Basin, Montney, and Bakken Shale plays. The simulator uses several empirical coefficients. We identify coefficients that have consistent values in the calibrations to all datasets. On the other hand, a few of the coefficients exhibit variability from dataset to dataset. We show how these parameters can be calibrated on a case-by-case basis prior to using the simulator for design optimization. Based on these case studies, we identified several opportunities to improve the simulator physics—by accounting for perforation ‘inline’ effects, including random variance in erosion coefficient, and increasing the amount of proppant suspension. Comparison across datasets shows that there is not a single consistent trend in heel-side or toe-side erosion bias. Different physical processes have opposing effects on heel/toe-side bias, and depending on the stage design and practical conditions, these processes can have different relative magnitudes. Correspondingly, the optimal perforation design varies from case-to-case, depending on which type of ‘bias’ is observed in the base design. The simulator predicts that measured erosion uniformity should be lower than the proppant or slurry uniformity. This result is supported by observations from a Bakken dataset in this study, where fiber-based slurry allocations yielded a significantly higher uniformity index than downhole imaging-based measurements. The implication is that the ‘uniformity index’ of erosion, observed from downhole imaging, cannot be taken as a direct proxy for the uniformity of proppant or fluid outflow. Finally, the simulator was applied to each field dataset to identify the optimal perforation design. The optimization procedure identified opportunities to improve the cluster-level uniformity index of proppant placement by a range of 0.12–0.19. Because each dataset requires case-specific calibration prior to optimization, there is no single ‘best’ design for all circumstances.
