|Theme||Visible||Selectable||Appearance||Zoom Range (now: 0)|
Suarez-Rivera, Roberto (W. D. Von Gonten Laboratories) | Panse, Rohit (W. D. Von Gonten Laboratories) | Sovizi, Javad (Baker Hughes) | Dontsov, Egor (ResFrac Corporation) | LaReau, Heather (BP America Production Company, BPx Energy Inc.) | Suter, Kirke (BP America Production Company, BPx Energy Inc.) | Blose, Matthew (BP America Production Company, BPx Energy Inc.) | Hailu, Thomas (BP America Production Company, BPx Energy Inc.) | Koontz, Kyle (BP America Production Company, BPx Energy Inc.)
Abstract Predicting fracture behavior is important for well placement design and for optimizing multi-well development production. This requires the use of fracturing models that are calibrated to represent field measurements. However, because hydraulic fracture models include complex physics and uncertainties and have many variables defining these, the problem of calibrating modeling results with field responses is ill-posed. There are more model variables than can be changed than field observations to constrain these. It is always possible to find a calibrated model that reproduces the field data. However, the model is not unique and multiple matching solutions exist. The objective and scope of this work is to define a workflow for constraining these solutions and obtaining a more representative model for forecasting and optimization. We used field data from a multi-pad project in the Delaware play, with actual pump schedules, frac sequence, and time delays as used in the field, for all stages and all wells. We constructed a hydraulic fracturing model using high-confidence rock properties data and calibrated the model to field stimulation treatment data varying the two model variables with highest uncertainty: tectonic strain and average leak-off coefficient, while keeping all other model variables fixed. By reducing the number of adjusting model variables for calibration, we significantly lower the potential for over-fitting. Using an ultra-fast hydraulic fracturing simulator, we solved a global optimization problem to minimize the mismatch between the ISIPs and treatment pressures measured in the field and simulated by the model, for all the stages and all wells. This workflow helps us match the dominant ISIP trends in the field data and delivers higher confidence predictions in the regional stress. However, the uncertainty in the fracture geometry is still large. We also compared these results with traditional workflows that rely on selecting representative stages for calibration to field data. Results show that our workflow defines a better global optimum that best represents the behavior of all stages on all wells, and allows us to provide higher-confidence predictions of fracturing results for subsequent pads. We then used this higher confidence model to conduct sensitivity analysis for improving the well placement in subsequent pads and compared the results of the model predictions with the actual pad results.
Abstract The proliferation of highly concentrated brine drilling fluids systems due to their enhanced performance benefits has instigated a plethora of technical studies on the mechanisms and control of their induced corrosion on downhole drilling tools and tubulars. The majority of these studies often overlook the effect of drill solids on corrosion rates. Therefore, a pragmatic and experimental study was conducted to investigate the effects of various factors on the corrosion rates of downhole tubulars with a streamlined focus on the obscure role of the understudied drill solids; which have not been fully elucidated. Drill pipe corrosion coupons and drilling fluids/solids obtained from 5 similar wells (located in Grande Prairie, Alberta, Canada) were utilized for experimental analysis. Wells 1 to 4 were on the same pad (surface drilling location) drilling the same formation with the same fluid properties, while the 5th well was on a different pad but drilled the same formation with the same fluid properties to exclude disparity. Industry-standard measurement was carried out on the live used corrosion coupon rings, drilling fluids and solids obtained from these wells to determine selected properties. The total solids content analysis was carried out using an OFITE API (American Petroleum Institute) filter press. Weight loss tests on drill pipe corrosion coupons were used to determine field corrosion rates which were bolstered with the Parr Hastelloy autoclave test in the Laboratory. The oxygen content was monitored using Hach 2100Q dissolved oxygen meter. Field data, images and experimental results showed that a rapid and minuscule increase of drill solids (as little as 1% v/v) in the active system can impact corrosion rates greater than chemical additives and even oxygen content. It was discovered that low concentration of solids can produce significant damage and a high corrosion potential in non-viscosified fluids thereby making live monitoring of drilling fluids’ properties a priority to mitigate corrosion. This study fills an important technical gap in corrosion study that is indispensable for the optimization of corrosion control in drilling operations. By carrying out a controlled and investigative study backed up with drilling field data and images, the effects of the less understood drill solids have been partially demystified.
To efficiently develop and operate a petroleum reservoir, it is important to have a model. Currently, numerical reservoir simulation is the accepted and widely used technology for this purpose. Data-driven reservoir modeling (also known as top-down modeling or TDM) is an alternative or a complement to numerical simulation. TDM uses the "big data" solutions of machine learning and data mining to develop--train, calibrate, and validate--full-field reservoir models based on measurements rather than mathematical formulation of our current understanding of the physics of fluid flow through porous media. Unlike other empirical technologies that forecast production such as decline curve analysis, or only use production/injection data for analysis (capacitance resistance model), TDM integrates all available field measurements such as well locations and trajectories, completions and stimulations, well logs, core data, well tests, seismic, as well as production/injection history, including wellhead pressure and choke setting.
Models are needed to develop and operate petroleum reservoirs efficiently. Data-driven reservoir modeling [(also known as top-down modeling (TDM)] is an alternative or a complement to numerical simulation. TDM uses the so-called "big-data" solution (machine learning and data mining) to develop (train, calibrate, and validate) full-field reservoir models on the basis of measurements rather than solutions of governing equations. Unlike other empirical technologies that forecast production, or only use production or injection data for its analysis, TDM integrates all available field measurements (well locations and trajectories, completions, stimulations, well logs, core data, well tests, seismic, and production/injection history--e.g., choke settings) into a full-field reservoir model by use of artificial-intelligence technologies. Intelligent Solutions, as the inventor of TDM, has recently released software application "IMagine" for TDM development. TDM is a full-field model wherein production [including gas/oil ratio (GOR) and water cut] is conditioned to all measured reservoir characteristics and operational constraints.
Domain experts who have a good understanding of artificial intelligence (AI) and machine learning (ML)--and have become expert practitioners of this technology--can be successful in modeling physics- and engineering-related problems purely based on data. When the data used for such modeling is generated using mathematical equations, the data-driven AI-based model is called a "smart proxy." When the data used for such modeling is generated using field measurements, the data-driven AI-based model is called a "data-driven model." Combining these two techniques is not a good idea. Here is the reason: When data-driven models are developed on the basis of field measurements, ML algorithms are trained to model the physics of the phenomena that are of interest.
Traditional statistics has been around for more than a century. Actually, the term was coined in Germany in 1749. If its connection with probability theory (randomness) is taken into account, then its history may even go as far back as the 16th century. Nevertheless, the point is that, unlike artificial intelligence (AI) and machine learning (ML), traditional statistics is not a new technology. In order to develop a better understanding of the fundamental differences between these two technologies, one should start with the seminal paper written by Leo Breiman, a well-known professor of statistics from University of Berkeley (Statistical Modeling: The Two Cultures).
Oil and gas exploration is facing an ever-increasing demand for cost-efficient drilling operations. Improvement of the rate of penetration (ROP) of the drill bit is key in solving the aforementioned challenge. The objective of this study is to develop a more accurate and effective predictive and optimization model for ROP that utilizes a hybrid artificial intelligence model based on an improved genetic algorithm (IGA) and artificial neural network (ANN) for further optimization of drilling processes.
Real field drilling datasets such as the bit type, bit drilling time, rotation per minute, weight on bit, torque, formation type, rock properties, hydraulics, and drilling mud properties are collected and input to train, validate and test the developed IGA-ANN model for ROP prediction and optimization. We apply a Savitzky-Golay (SG) smoothing filter to reduce the noise from the raw datasets. We apply IGA to find the optimal structures, parameters, and types of input of the ANN. By using supplementary population, multi-type crossover and mutation and adaptive dynamics probability adjustments, the developed model, IGA-ANN, avoid the limited optimization and local convergence problems in the classical Genetic Algorithm (GA). Using the developed prediction model, we obtain the optimal operational parameter within a region considering drilling equipment capability and wear to maximize ROP.
From numerical results, we find that the optimal structures and parameters of the ANN can be obtained efficiently by the developed method. For comparison, we compare IGA-ANN with the classical wrapper algorithm for parameter selection. The results indicate that IGA-ANN is more stable and accurate than the wrapper algorithm. We compare the true ROP and predicted ROP from the developed IGA-ANN model using accuracy indicators such as root mean square error, mean absolute error, and regression coefficient (R2). For comparison, the accuracy of the classical regression model is presented. We find that IGA-ANN yielded more accurate test results (R2 = 0.97). We compare the results of IGA-ANN trained using SG smoothing filter processed data and raw data. The test results show that the noise reduction approach used is very efficient in increasing the accuracy of IGA-ANN. Using the developed model, we optimize the choice of drilling operational parameters within a region considering drilling equipment capability and wear. We find that the optimization can increase average ROP significantly.
We develop an efficient and robust algorithm, IGA-ANN, for ROP prediction and optimization. Compared with the classical wrapper algorithm and multiple regression model, the IGA-ANN can efficiently optimize the structures, parameters and types of inputs of ANN to achieve higher ROP prediction accuracy. By utilizing the developed model, we can efficiently maximize ROP and minimize the drilling operation cost.
Masoudi, Rahim (Petronas MPM) | Mohaghegh, Shahab D. (West Virginia University) | Yingling, Daniel (Intelligent Solutions, Inc.) | Ansari, Amir (Intelligent Solutions, Inc.) | Amat, Hadi (Petronas MPM) | Mohamad, Nis (Petronas COE) | Sabzabadi, Ali (Petronas MPM) | Mandel, Dipak (Petronas COE)
Using commercial numerical reservoir simulators to build a full field reservoir model and simultaneously history match multiple dynamic variables for a highly complex, offshore mature field in Malaysia, had proven to be challenging, manpower intensive, highly expensive, and not very successful. This field includes almost two hundred wells that have been completed in more than 60 different, non-continuous reservoir layers. The field has been producing oil, gas and water for decades. The objective of this article is to demonstrate how Artificial Intelligence (AI) and Machine Learning is used to build a purely data-driven reservoir simulation model that successfully history match all the dynamic variables for all the wells in this field and subsequently used for production forecast. The model has been validated in space and time.
The AI and Machine Learning technology that was used to build the dynamic reservoir simulation and modeling is called spatio-temporal learning. Spatio-temporal learning is a machine-learning algorithm specifically developed for data-driven modeling of the physics of fluid flow through porous media. Spatio-temporal learning is used in the context of Deconvolutional Neural Networks. In this article Spatio-temporal Learning and Deconvolutional Neural Networks will be explained. This new technology is the result of more than 20 years of research and development in the application of AI and Machine Learning in reservoir modeling. This technology develops a coupled reservoir and wellbore model that for this particular oil & gas field in Malaysia uses choke setting, well-head pressure and well-head temperature as input and simultaneously history matches Oil production, GOR, WC, reservoir pressure, and water saturation for more than a hundred wells through a completely automated process.
Once the data-driven reservoir model is developed and history matched, it is blind validated in space and time in order to establish a reliable and robust reservoir model to be used for decision making purposes and opportunity generation to maximise the field value. The concepts and the methodology of history match of multiple wells, individual offshore platforms, and the entire field will be presented in this article along with the results of blind validation and production forecasting. Results of using this model to perform uncertainty quantification will also be presented.
A case study of a highly complex mature field with large number of wells and years of production has been used to be studied and simulated by this data-driven approach. Time, efforts, and resources required for the development of the dynamic reservoir simulation models using AI and Machine Learning is considerably less than time and resources required using the commercial numerical simulators. It is validated that the TDM developed model can make very reasonable prediction of field behavior and production from the existing wells based on modification of operational constraints and can be a prudent complementary tool to conventional numerical simulators for such complex assets.
ABSTRACT Mining induced stress redistribution in underground coal mines plays a significant role in pillar and support system design, hence in the safety of mining operations. Western U.S. longwall coal mines usually apply two-entry yield pillar system, since the overburden in the Western US longwall mines are often deeper and stronger than the Eastern US mines. For this reason, it is critical to approximate the stress redistribution along the yield pillar during the longwall retreat for optimizing the support system. To investigate the stress redistribution, a numerical-model-based approach was used in a 1,500 ft deep longwall mine (Mine A) from the Western US. In this modeling approach, a systematic procedure is used to estimate the model's input parameters. It was found that the default input parameters, which have been successfully applied in the Eastern US mines, over estimates the surface subsidence and gob loads for the Western US overburden geology and are not applicable in that region. The response of the gob is calibrated with the back analysis of subsidence data from Mine A. The calibrated model results were verified with the subsidence measurement from the same mine. The model results demonstrate the surface subsidence due to longwall mining can be satisfactorily simulated with this updated modeling approach. 1. INTRODUCTION AND BACKGROUND INFORMATION Mining induced stress redistribution in an underground coal mine plays a significant role in pillar and support system design, hence in the safety of mining operations (Campoli, 2015). Even though the reported roof fall rates in longwall tailgates have been decreasing in the past few years, roof falls are still a safety hazard for underground coal miners (Sears, et al., 2019). In 2017, 91 lost time injuries occurred due to the roof fall accidents in the US underground coal mines, and additional 48 roof falls occurred with no lost days (MSHA, 2018). In addition to the serious ground fall hazards, large gateroad deformation can also disrupt the ventilation system, block the escape ways, and increase the potential for methane accumulation (MSHA, 2010).
Agyei-Dwarko, N. Y. (Norwegian Geotechnical Institute) | Kveldsvik, V. (Norwegian Geotechnical Institute) | Chiu, J. K. Y. (Norwegian Geotechnical Institute) | Smebye, H. (Norwegian Geotechnical Institute)
This study presents the use of imagery from Unmanned Aerial Vehicle (UAV) as a primary means of structural geological data collection, complimented by traditional methods to assess the stability of high rock slopes along the Rugtvedt-Dørdal section of the E-18 highway under construction in Norway. The road has about 10km of road cuts which are between 10 and 35m high, making traditional mapping methods tedious. High resolution images taken by UAV were used to construct 3D geological models of the rock masses using the Agisoft Metashape software from which structural data such as dip, dip direction, joint roughness, joint spacing, etc. were extracted by use of the PointStudio software by Maptek. The extracted parameters were found to be consistent with data collected by traditional mapping methods. The digital data was supplemented with field data on groundwater conditions, weathering and rock strength and served as the basis for detailed stability analysis of rock slopes along the road using the Rocscience software suite. This workflow is found to be efficient, reliable, and allowed for the collection of data over large areas in a rapid manner. This methodology as an excellent supplement to traditional field mapping methods in areas of steep and high slopes with low access.
The stability analysis of rock slopes relies on the accurate collection of structural geological and geomechanical data as rock slope stability is often structurally controlled. This is usually done by traditional mapping methods on outcrops and slopes. In large projects and in areas of rugged and high topography with limited access, this task is often time-consuming and sometimes completely impractical. In recent years, several attempts and advances have been made in the use of photogrammetry as a tool in the collection of such data at various scales. In this study, we present the systematic collection of structural data from 3D models derived from high resolution drone imagery compared with data from traditional methods to test the viability of this method on an active construction project and to form the basis of further stability assessments.