The field-scale design of chemical enhanced oil recovery (cEOR) processes requires running complex numerical models that are computationally demanding. This paper provides an efficient screening platform for the cEOR feasibility study by presenting five artificial neural network (ANN) based models. We constructed 1,100 ANN training cases using CMG-STARS to capture the variation in reservoir petrophysical properties and the range of injected chemicals properties for a five-spot pattern. The design parameters were coupled with the reservoir properties using several functional links to optimize the ANN models and improve their performances. The training cases were employed using back-propagation methods to construct one forward model (Model #1) and four inverse models. Model #1 predicts reservoir response (i.e., oil rate, water cut, injector bottomhole pressure, cumulative oil) for known reservoir characteristics (i.e., permeability, thickness, residual oil saturation, chemical adsorption) and project design parameters (i.e., pattern size, chemical slug size and concentration), Model #2 predicts reservoir characteristics by history matching the reservoir response, and Model #3 predicts project design parameters for known reservoir response and characteristics. Models #4 and #5 predict project design parameters for a targeted cumulative oil volume and project duration time, which is useful for economical evaluation before the implementation of cEOR projects.
The validation results show that the developed ANN-based models closely predict the numerical results. In addition, the models are able to reduce the computational time by four orders of magnitude, which is significant considering the complexity of cEOR modeling and the need for reliable and efficient tools in building cEOR feasibility studies. In terms of accuracy, Model #1 has a prediction error of 5% whereas the error for other four inverse ANN models is about 20–40%. To enhance the performance of the inverse ANN models, we changed the ANN structure, increased training cases, and used functional links, which slightly reduced the error. Further, we introduced a back-check loop that uses the predicted parameters from the inverse ANN models as inputs in the forward ANN model. A comparison of back-check results for the reservoir response with the numerical results delivers a relatively small error of 10%, revealing the non-uniqueness of solutions obtained from the inverse ANN models.
You, Junyu (Petoleum Recovery Research Center) | Ampomah, William (Petoleum Recovery Research Center) | Kutsienyo, Eusebius Junior (Petoleum Recovery Research Center) | Sun, Qian (Petoleum Recovery Research Center) | Balch, Robert Scott (Petoleum Recovery Research Center) | Aggrey, Wilberforce Nkrumah (KNUST) | Cather, Martha (Petoleum Recovery Research Center)
This paper presents an optimization methodology on field-scale numerical compositional simulations of CO2 storage and production performance in the Pennsylvanian Upper Morrow sandstone reservoir in the Farnsworth Unit (FWU), Ochiltree County, Texas. This work develops an improved framework that combines hybridized machine learning algorithms for reduced order modeling and optimization techniques to co-optimize field performance and CO2 storage.
The model's framework incorporates geological, geophysical, and engineering data. We calibrated the model with the performance history of an active CO2 flood data to attain a successful history matched model. Uncertain parameters such as reservoir rock properties and relative permeability exponents were adjusted to incorporate potential changes in wettability in our history matched model.
To optimize the objective function which incorporates parameters such as oil recovery factor, CO2 storage and net present value, a proxy model was generated with hybridized multi-layer and radial basis function (RBF) Neural Network methods. To obtain a reliable and robust proxy, the proxy underwent a series of training and calibration runs, an iterative process, until the proxy model reached the specified validation criteria. Once an accepted proxy was realized, hybrid evolutionary and machine learning optimization algorithms were utilized to attain an optimum solution for pre-defined objective function. The uncertain variables and/or control variables used for the optimization study included, gas oil ratio, water alternating gas (WAG) cycle, production rates, bottom hole pressure of producers and injectors. CO2 purchased volume, and recycled gas volume in addition to placement of new infill wells were also considered in the modelling process.
The results from the sensitivity analysis reflect impacts of the control variables on the optimum results. The predictive study suggests that it is possible to develop a robust machine learning optimization algorithm that is reliable for optimizing a developmental strategy to maximize both oil production and storage of CO2 in aqueous-gaseous-mineral phases within the FWU.
Japan Agency for Marine-Earth Science and Technology (JAMSTEC) operates the scientific deep-sea drilling vessel Chikyu for scientific studies on past climate changes, long-term crust transportation, biological and abiotic processes associated with hydrocarbon production, and earthquakes, the last being one of the most important research topics. Chikyu conducted the Japan Trench Fast Drilling Program (JFAST) to obtain core samples from sediment layers under the seabed for study on the large Tohoku earthquake which triggered the devastating tsunami in 2011. Chikyu has also started to conduct the Nankai Trough Seismogenic Zone Experiment (NanTroSEIZE) as the Nankai Trough is one of the most active earthquake zones. As one of the primary goals of scientific drilling is to evaluate sediment properties, it is highly beneficial to characterize the lithology of the drilling layer during drilling operations. Even an approximation of these properties could potentially provide valuable information for conducting coring operations. Thus, a previous study attempted to discuss the properties of sediment layers using surface drilling data, and estimated the shear stress of the sediment. In this study, we summarize the method applied in the previous paper for estimating the torque at the drill bit from the surface measured drilling data for NanTroSEIZE deep riser drilling operation. This data was then used to directly calculate the shear stress of the sediment. The drilling torque at the bit differed from the surface torque, where this difference increased with increasing drilling depth. Our estimations of the drilling torque were validated using torque data obtained from logging while drilling (LWD) operations. We also used machine-learning approaches to predict the lithology, where learning data was created from surface drilling data and lithology information from core samples obtained during past scientific drilling operations. Machine learning was then applied using neural network algorithms by tuning the number of layers to create a predictive model. This paper discusses the preliminary attempt to predict the lithology using machine-learning approaches for NanTroSEIZE and JFAST data.
Hadi, Farqad (Petroleum Engineering Department, Baghdad University) | Albehadili, Ali (Iraqi Drilling Company) | Jassim, Abduihussein (Najaf Oil Fields) | Almahdawi, Faleh (Petroleum Engineering Department, Baghdad University)
Formulating a prediction tool that can estimate the formation permeability in uncored wells is of particular importance for many applications related to reservoir simulation and production management. Although formation permeability can be obtained from a laboratory or from a reservoir, core analysis and well-test data are limited due to cost and time-saving purposes. A major challenge of previous methods is that they are required other parameters to be previously computed such as porosity and water saturation. In addition, they are affected by the uncertainty that introduced by the cementation factor and saturation exponent. This study presents two prediction methods, multiple regression analysis (MRA) and artificial neural networks (ANNs), to estimate formation permeability using conventional well log data.
The prediction methods were demonstrated by means of a field case in SE Iraq. The study uses core/well log data from Mishrif reservoir which is mainly composed of carbonate (limestone) formations. Two traditional methods were reviewed and presented for permeability determination. These methods are the classical method and the flow zone indicator (FZI) method.
At the same porosity, the results showed a wide range of formation permeability prediction. This result gives a special attention to the assumption that the relationship between permeability and porosity is generally unique in carbonate environments. The deep lateral log resistivity appears to be more conservative in the permeability function rather than other parameters, followed in decreasing order by bulk density, sonic travel time, micro and shallow resistivities, and shale volume. Although the presented models based on RA and ANNs resemble to be closely in determining the formation permeability, the correlation coefficient of ANNs was found to be higher than that obtained from RA, which indicated that the ANNs is more precise than RA. The comparison among previous methods shows the superiority of the FZI method rather than the classical method. However, core porosity and permeability should be previously determined to apply FZI method. This study presents efficient and cost-effective models for a prediction of permeability in uncored wells by incorporating conventional well logs.
In this study, the authors investigated a fully data-driven approach using artificial neural networks (ANNs) for real-time virtual flowmetering and back-allocation in production wells. The authors present a new data-driven approach to estimate the injection rate in all noninstrumented wells in a large waterflooding operation accurately.
In this study, the authors investigated a fully data-driven approach using artificial neural networks (ANNs) for real-time virtual flowmetering and back-allocation in production wells. Australian technology developer MezurX is touting its newly introduced flow, density, and mud monitoring system as a significantly better alternative to the widely used Coriolis meter. Technology developers are working on a new ultrasonic flowmeter for drilling fluids aimed at obtaining better measurements than current systems with far fewer disruptions to rig systems.
AGI stands for artificial general intelligence, a hypothetical computer program that can perform intellectual tasks as well as, or better than, a human. AGI will make today’s most advanced AIs look like pocket calculators. In collaboration with Stanford University and Brown University, Google explores how existing knowledge in an organization can be used as noisier, higher-level supervision—or, as it is often termed, weak supervision—to quickly label large training data sets. Robots have been a part of industrial production for decades, but the interface between humans and robots has changed as automation technologies increased in complexity, scope, and scale. Once a novelty, collaborative robots are projected to become a significant element of the automation landscape.
New funding for a chatbot technology, or smart assistant, represents the latest development in the Norwegian operator’s drive toward digitalization. This digital deal is helping to make augmented reality a new reality for oil and gas operations. This paper demonstrates how engineers can take advantage of their most-detailed completions and geomechanical data by identifying trends arising from past detailed treatment analyses. Dubbed the technology of the decade, AI has been the catchphrase on every futurist’s tongue. From customer support chatbots to smart assistants, AI has begun to transform numerous industry verticals.
Alkinani, Husam H. (Missouri University of Science and Technology) | Al-Hameedi, Abo Taleb T. (Missouri University of Science and Technology) | Dunn-Norman, Shari (Missouri University of Science and Technology) | Alkhamis, Mohammed M. (Missouri University of Science and Technology) | Mutar, Rusul A. (Ministry of Communications and Technology)
Lost circulation is a complicated problem to be predicted with conventional statistical tools. As the drilling environment is getting more complicated nowadays, more advanced techniques such as artificial neural networks (ANNs) are required to help to estimate mud losses prior to drilling. The aim of this work is to estimate mud losses for induced fractures formations prior to drilling to assist the drilling personnel in preparing remedies for this problem prior to entering the losses zone. Once the severity of losses is known, the key drilling parameters can be adjusted to avoid or at least mitigate losses as a proactive approach.
Lost circulation data were extracted from over 1500 wells drilled worldwide. The data were divided into three sets; training, validation, and testing datasets. 60% of the data are used for training, 20% for validation, and 20% for testing. Any ANN consists of the following layers, the input layer, hidden layer(s), and the output layer. A determination of the optimum number of hidden layers and the number of neurons in each hidden layer is required to have the best estimation, this is done using the mean square of error (MSE). A supervised ANNs was created for induced fractures formations. A decision was made to have one hidden layer in the network with ten neurons in the hidden layer. Since there are many training algorithms to choose from, it was necessary to choose the best algorithm for this specific data set. Ten different training algorithms were tested, the Levenberg-Marquardt (LM) algorithm was chosen since it gave the lowest MSE and it had the highest R-squared. The final results showed that the supervised ANN has the ability to predict lost circulation with an overall R-squared of 0.925 for induced fractures formations. This is a very good estimation that will help the drilling personnel prepare remedies before entering the losses zone as well as adjusting the key drilling parameters to avoid or at least mitigate losses as a proactive approach. This ANN can be used globally for any induced fractures formations that are suffering from the lost circulation problem to estimate mud losses.
As the demand for energy increases, the drilling process is becoming more challenging. Thus, more advanced tools such as ANNs are required to better tackle these problems. The ANN built in this paper can be adapted to commercial software that predicts lost circulation for any induced fractures formations globally.
Computer vision (CV) techniques were applied to X-ray computed tomographic (CT) images of reservoir cores to evaluate their potential for rapidly identifying fractures and other lithologic/geologic characteristics. The analyses utilized feature-labeled CT cross sectional images, themselves distributed submillimeter voxels of density- and atomic number-sensitive CT Numbers, as inputs to a Fully Convolutional Neural Network (FCN) for semantic segmentation of reservoir core features. In FCN, an image is interrogated using a series of sliding windows at various scales to create weighted filters to reduce error between classes in training images. These networks of filter layers were used to assign probabilities of classes, which were upscaled back to the original image dimensions resulting in probabilistic class assignments onto each pixel. FCN model accuracy, defined by its ability to replicate manually-assigned labels in the raw (unannotated) training image stack, was at least 80% and generally improved with the size of the training set. Once the labels were assigned, the underlying feature frequency, orientation, and size were measured in 3D volume reconstructions using algorithms modified from standard image analysis software. This method allowed users to endow a classification model with subject matter knowledge for further, autonomous label prediction. Thus, while initial image annotation was labor intensive, subsequent images were rapidly classified once the model was built. The classified labels were analyzed for abundance, orientation, and size of fractures were calculated to characterize spatial information of these features. FCN combined with fracture labeling improved knowledge capture and automation of fracture identification. Models trained by high quality 3D datasets can greatly reduce the time needed to describe subsequent core. The method demonstrated is not limited to fractures, other lithologic/geologic features could be trained using the same method, which may result in additional efficiencies.