Tight gas sandstone reservoirs vary widely in terms of rock type, depositional environment, mineralogy and petrophysical properties. For this reason, estimating their permeability is a challenge when core is not available because it is a property that cannot be measured directly from wire-line logs. The aim of this work is to create an automatic tool for rock microstructure classification as a first step for future permeability prediction. Permeability can be estimated from porosity measured using wire-line data such as derived from density-neutron tools. However, without additional information this is highly inaccurate because porosity-permeability relationships are controlled by the microstructure of samples and permeability can vary by over five orders of magnitude. Experts can broadly estimate porosity-permeability relationships by analysing the microstructure of rocks using Scanning Electron Microscopy (SEM) or optical microscopy. Such estimates are, however, subjective and require many years of experience. A Machine Learning model for the automation of rock microstructure determination on tight gas sandstones has been built using Convolutional Neural Networks (CNNs) and trained on backscattered images from cuttings. Current results were obtained by training the model on around 24,000 Back Scattering Electron Microscopy (BSEM) images from 25 different rock samples. The obtained model performance for the current dataset are 97% of average of both validation and test categorical accuracy. Also, loss of 0.09 and 0.089 were obtained for validation and test correspondingly. Such high accuracy and low loss indicate an overall great model performance. Other metrics and debugging techniques such Gradient-weighted Class Activation Mapping (Grad-CAM), Receiver Operator Characteristics (ROC) and Area Under the Curve (AUC) were considered for the model performance evaluation obtaining positive results. Nevertheless, this can be improved by obtaining images from new already available samples and make the model generalizes better. Current results indicate that CNNs are a powerful tool and their application over thin section images is an answer for image analysis and classification problems. The use of this classifier removes the subjectivity of estimating porosity-permeability relationships from microstructure and can be used by non-experts. The current results also open the possibility of a data driven permeability prediction based on rock microstructure and porosity from well logs.

The execution phase of the wells technical assurance process is a critical procedure where the drilling operation commences and the well planning program is implemented. During drilling operations, the real-time drilling data are streamed to a real-time centre where it is constantly monitored by a dedicated team of monitoring specialists. If any potential issues or possible opportunities arise, the team will communicate with the operation team on rig for an intervention. This workflow is further enhanced by digital initiatives via big data analytics implementation in PETRONAS. The Digital Standing Instruction to Driller (Digital SID) is a drilling operational procedures documentation tool meant to improve the current process by digitalizing information exchange between office and rig site. Boasting multi-operation usage, it is made fit to context and despite its automated generation, this tool allows flexibility for the operation team to customize the content and more importantly, monitor the execution in real-time. Another tool used in the real-time monitoring platform is the dynamic monitoring drilling system where it allows real-time drilling data to be more intuitive and gives the benefit of foresight. The dynamic nature of the system means that it will update existing roadmaps with extensive real-time data as they come in, hence improving its accuracy as we drill further. Furthermore, an automated drilling key performance indicator (KPI) and performance benchmarking system measures drilling performance to uncover areas of improvement. This will serve as the benchmark for further optimization. On top of that, an artificial intelligence (AI) driven Wells Augmented Stuck Pipe Indicator (WASP) is deployed in the real-time monitoring platform to improve the capability of monitoring specialists to identify stuck pipe symptoms way earlier before the occurrence of the incident. This proactive approach is an improvement to the current process workflow which is less timely and possibly missing the intervention opportunity. These four tools are integrated seamlessly with the real-time monitoring platform hence improving the project management efficiency during the execution phase. The tools are envisioned to offer an agile and efficient process workflow by integrating and tapering down multiple applications in different environments into a single web-based platform which enables better collaboration and faster decision making.

The B Field is located in the South China Sea, about 45 KM offshore Sarawak, Malaysia, in a water depth approximately 230 ft. Its structure is generally regarded as a gentle rollover anticline with collapsed crest resulting from growth faulting. The reservoirs were deposited in a coastal to shallow marine with some channels observed. Multiple stacked reservoirs consist of a series of very thick stacked alternating sandstone and minor shale layers with differing reservoir properties. The shallow zones are unconsolidated, and the wells were completed with internal gravel packs. Wells in B Field mostly were completed in multi-layered reservoirs as dual strings with SSDs and meant to produce as a commingled production. The well BX is located within B Field and designed as oil producer well with a conventional tubing jointedElectrical Submersible Pump (ESP) system which was installed back in 2008. Refer to figure 1 , the initial completion schematic is 3-1/2″ single string that consist of the single production packer, gas lift mandrel, tubing retrievable Surface Controlled Subsurface Safety Valve (SCSSV) and ESP. The production packers equipped with the feed thru design to accommodate the ESP cable and the gas vent valve as part of the ESP completion design. The gas lift mandrel was installed in the completion string as a backup artificial lift method to receive the gas lift and orifice valve in the event of the conventional ESP failed. Hence the well still able to produce by introducing the gas thru the annulus to activate the gas lift valve. Eventually throughout the end of the the field life, the well would depend on the ESP system for the primary lifting method due to gas lift depth limitation and the gas supply. The conventional ESP failed after seven years of operation which is above the average ESP lifetime. The well last produced at a flow rate with 28 % water cut, however the well is not at the end of the field life. Based on the economical study with the right technology and cost efficient approach, the well still economicaly profitable. The Thru Tubing (TT) ESP technology is approached as cost effective solution compare to fully well workover. Despite a couple of operational challenges, for example, setting the cable hanger, maintaining downhole barrier requirement, the Thru Tubing Electrical Submersible Pump Cable Deployed (TTESP CD) and Cable Thru Insert Safety Valve (CT-ISV) was successfully installed. Several post-installation findings have uncovered some problems which are requiring some additional technical and operation improvement for future similar applications. This paper will highlight the deployment of the Cable Thru Insert Safety Valve (CT-ISV) that was successfully installed as pilot, which is the first application in the world, and also highlights the success, lesson learnt and improvement for future requirement for the CT-ISV application as one of the solution for retrofitting completion application without jeopardizing the well integrity. This achievement is collaboration between Company and service partner as the technology and deployment under the proprietary scope. Further technology application, the replication of this insert safety valve was conducted and successfully deployed on other three wells.

Fault detection is an important, but time-consuming task in seismic data interpretation. Traditionally, seismic attributes, such as coherency ( Marfurt et al., 1998 ) and curvature ( Al-Dossary et al., 2006 ) are used to detect faults. Recently, machine learning methods, such as convolution neural networks (CNNs) are used to detect faults, by applying various semantic segmentation algorithms to the seismic data ( Wu et al., 2019 ). The most used algorithm is U-Net ( Ronneberger et al., 2015 ), which can accurately and efficiently provide probability maps of faults. However, probabilities of faults generated by semantic segmentation algorithms are not sufficient for direct recognition of fault types and reconstruction of fault surfaces. To address this problem, we propose, for the first time, a workflow to use instance segmentation algorithm to detect different fault lines. Specifically, a modified CNN ( LaneNet; Neven et al., 2018 ) is trained using automatically generated synthetic seismic images and corresponding labels. We then test the trained CNN using both synthetic and field collected seismic data. Results indicate that the proposed workflow is accurate and effective at detecting faults.

Stuck pipe is one of the leading causes of non-productive time (NPT) while drilling. Machine learning (ML) techniques can be used to predict and avoid stuck pipe issues. In this paper, a model based on ML to predict and prevent stuck pipe related to differential sticking (DS) is presented. The stuck pipe indicator is established by detecting and predicting abnormalities in the drag signatures during tripping and drilling activities. The solution focuses on detecting differential sticking risk via assessing hookload signatures, based on previous experience from historical wells. Therefore, selecting the proper training set has proven to be a crucial stage of model development, especially considering the challenges in data quality. The model is trained with historical wells with and without differential sticking issues. The solution is based on the Artificial Neural Network (ANN) approach. The model is designed to provide users, i.e., driller or monitoring specialist, a warning whenever a risk is identified. Since multi-step forecasting is used, the warning is given with enough time for the driller or monitoring specialist to evaluate which preventative action or intervention is necessary. The warnings are provided typically between 30 minutes and 4 hours ahead. The model validation includes the performance metrics and a confusion matrix. Practical cases with real-time wells are also provided. The ML model was proven robust and practical with our data sets, for both historical and live wells. The huge amount of data produced while drilling holds valuable information and when smartly fed into an Artificial Intelligence (AI) model, it can prevent NPT such as stuck pipe events as demonstrated in this paper.

Conventionally, a bit is selected from offset well bit run summaries. This method of selection is not always accurate since each bit is run under different conditions which might not be reflected in an offset study analysis. The large quantities of data generated from real time measurements in offset wells makes machine learning the ideal tool for analysis and comparison. Artificial Neural Network (ANN) is a relatively simple machine learning tool that combines inputs and calculation layers to compute a specified output layer. The ANN is fed over thousands of data points from 17-1/2 in hole sections across multiple wells. A specific model is then trained for every bit with weight on bit (WOB), rotary speed (RPM), bit hydraulics, and lithological properties as inputs and rate of penetration (ROP) as output. The model is finalized when a satisfactory statistical set of KPI's are achieved. Using a combination of Monte-Carlo analysis and sensitivity analysis, different bits are compared by varying parameters for the same bit and varying the bit under the same parameters. A bit and its optimized parameters are proposed, resulting in an average instantaneous ROP improvement of 32%. Performance benchmarked with individual drilling parameters shows improved ROP response to WOB, RPM, and bit hydraulics in the optimized run. This project solidifies machine learning as a powerful tool for bit selection and parameter optimization to improve drilling performance. Machine learning will become a significant part of well planning, design, and operations in the future. This study demonstrates how ANN's can be used to learn from previous operations and influence planning decisions to improve bit performance.

The technique of interwell tracer testing is considered as one of the most effective method to identify the thief zone (TZ) in reservoirs. However, in heavy oil reservoirs, tracer breakthrough curves are mostly parabolic and unimodal, thus resulting in slight differences between curves. It is inefficient and inaccurate to identify different types of curves with traditional methods applied to characterize the levels of TZs. In this paper, convolutional neural network (CNN) is applied to construct a classification model for the automatic identification of the levers of TZs. According to the TZs criteria specified on the field, the analytical tracer transport model was applied to generate 3000 curves as the sample, which can meet the requirements of model training accuracy. In the meantime, One-hot encoding, Xavier initialization, Adam optimizer, and mini-batch normalization were used to construct the model, and the key parameters are optimized to improve the performance of the model. The results show that the appropriate activation function is ReLU and the optimal dropout rate is 0.5. Moreover, the construction of CNN with discrete data points (DDP-CNN) as input contributed to a further improvement of classification accuracy of tracer curves. The accuracy of DDP-CNN in training set is 0.96, which is 14% and 23% higher than random forest (RF) and k-means, respectively. In practical applications, DDP-CNN proves capable to correctly classify 88 of the 100 curves.

Fluid properties are key factors for predicting single well productivity, well test interpretation and oilfield recovery prediction, which directly affect the success of ODP program design. The most accurate and direct method of acquisition is underground sampling. However, not every well has samples due to technical reasons such as excessive well deviation or high cost during the exploration stage. Therefore, analogies or empirical formulas have to be adopted to carry out research in many cases. But a large number of oilfield developments have shown that the errors caused by these methods are very large. Therefore, how to quickly and accurately obtain fluid physical properties is of great significance. In recent years, with the development and improvement of artificial intelligence or machine learning algorithms, their applications in the oilfield have become more and more extensive. This paper proposed a method for predicting crude oil physical properties based on machine learning algorithms. This method uses PVT data from nearly 100 wells in Bohai Oilfield. 75% of the data is used for training and learning to obtain the prediction model, and the remaining 25% is used for testing. Practice shows that the prediction results of the machine learning algorithm are very close to the actual data, with a very small error. Finally, this method was used to apply the preliminary plan design of the BZ29 oilfield which is a new oilfield. Especially for the unsampled sand bodies, the fluid physical properties prediction was carried out. It also compares the influence of the analogy method on the scheme, which provides potential and risk analysis for scheme design. This method will be applied in more oil fields in the Bohai Sea in the future and has important promotion value.

Subsurface engineers pivot on surveillance of reservoir performance for future production rates and plan the optimization strategies at earliest. There are some techniques preferred for unconventional reservoirs such as numerical simulation and decline curve analysis (DCA) for production forecasting, but the uncertainty of uneconomical well test data often occurs in unconventional resources. Moreover, reservoir engineers can also hit a tailback in optimizing and tuning the model. Further, for DCA this approach is only appropriate for well/reservoir that are under boundary dominant flow regime, whereas fracture dominant flow regime is often observed for a longer period in unconventional hydraulically fractured reservoirs. Therefore, to resolve this issue, oil & gas industry (O&G) can adopt AI (Artificial Intelligence) based Algorithms for production forecasting. This paper presents a data-driven algorithm, known as Artificial Neural Networks (ANNs), along with time series forecasting that is a well-known statistical technique. Machine learning model trained by a past well performance data such as tubing head pressure (THP), flowing bottom-hole pressure can predict future production rates. This can be an efficient technique for subsurface engineers to monitor and optimize well performance. Time series neural networks were used for training the model at top and bottom node of the well with variating pressures in the past. After training and validation, the model predicted a target parameter that was gas rate. ANNs are inspired by biological neurons that are present in human brain, a powerful computing tool to make decisions after fueling itself with data. Moreover, prediction (t+1) nonlinear automated regression is preferred for accurate step ahead. Production rates and constraints of unconventional reservoirs were used to train the model. In our results, the NN based model gave the co-efficient of determination (R 2 ) of 0.996 that shows nearly an exact precision. Furthermore, the values generated from NN Model and Arp's decline curve calculations were plotted for validation and it turned out that ANN can accurately predict the parameters. The Neural Network model is a novel approach for production forecasting, of unconventional reservoirs and help engineers in corporate decision making. This approach can mitigate the need of uneconomical well test operations and further provide confidence to production engineers in terms of data and result expectations.

Well real-time flow rate is one of the most important production parameters in oilfield and accurate flow rate information is crucial for production monitoring and optimization. With the wide application of permanent downhole gauge (PDG), the high-frequency and large volume of downhole temperature and pressure make applying of deep learning technique to predict flow rate possible. Flow rate of production well is predicted with long short-term memory (LSTM) network using downhole temperature and pressure production data. The specific parameters of LSTM neural network are given, as well as the methods of data preprocessing and neural network training. The developed model has been validated with two production wells in the Volve Oilfield, North Sea. The field application demonstrates that the deep learning is applicable for flow rate prediction in oilfields. LSTM has the better performance of flow rate prediction than other five machine learning methods, including support vector machine (SVM), linear regression, tree, and Gaussian process regression. The LSTM with a dropout layer has a better performance than a standard LSTM network. The optimal numbers of LSTM layers and hidden units can be adjusted to obtain the best prediction results, but more LSTM layers and hidden units lead to more time of training and prediction, and LSTM model might be unstable and cannot converge. Compared with only downhole pressure or temperature data used as input parameters, flow rate prediction with both of downhole pressure and temperature used as input parameters has the higher prediction accuracy.

Optimizing multiple assets under uncertain techno-economic conditions and tight government policies is challenging. Operator needs to establish flexible Plan of Development (POD)s and put priority in developing multiple fields. The complexity of production and the profit margin should be simultaneously evaluated. In this work, we present a new workflow to perform such a rigorous optimization under uncertainty using the case study of PHE ONWJ, Indonesia. We begin the workflow by identifying the uncertain parameters and their prior distributions. We classify the parameters into three main groups: operations-related (geological complexity, reserves, current recovery, surface facilities, and technologies), company-policies-related (future exploration plan, margin of profit, and oil/gas price), and government-related (taxes, incentives, and fiscal policies). A unique indexing technique is developed to allow numerical quantification and adapt with dynamic input. We then start the optimization process by constructing time-dependent surrogate model through training with Monte Carlo sampling. We then perform optimization under uncertainty with multiple scenarios. The objective function is the overall Net Present Value (NPV) obtained by developing multiple fields. This work emphasizes the importance of the use of time-dependent surrogate approach to account risk in the optimization process. The approach revises the prior distribution with narrow-variance distribution to make reliable decision. The Global Sensitivity Analysis (GSA) with Sobol decomposition on the posterior distribution and surrogate provides parameters’ ranking and list of heavy hitters. The first output from this workflow is the narrow-variance posterior distribution. This result helps to locate the sweet spots. By analyzing them, operator can address specific sectors, which are critical to the NPV. PHE ONWJ, as the biggest operator in Indonesia, has geologically scattered assets, therefore, this first output is essential. The second output is the list of heavy hitters from GSA. This list is a tool to cluster promising fields for future development and prioritize their development based on the impact towards NPV. Since all risks are carried by the operator under the current Gross Split Contract, this result is advantageous for decision-making process. We introduce a new approach to perform time-dependent, multi-asset optimization under uncertainty. This new workflow is impactful for operators to create robust decision after considering the associated risks.

G oilfield in China is a dual medium oilfield dominated by carbonate reservoir, where faults and fractures are well developed. After G oilfield was put into production, water cut has rised quickly and oil production decreased rapidly. The heterogeneity of fracture development is very strong. Because of unclear understanding of the fracture distribution, oilfield development plan can not accurately identify risks and potentials accurately. As a result, production goals are not as expected. Due to limited core data and imaging logging data, conventional methods cannot characterize fracture information in this oilfield accurately and comprehensively. In order to solve the problems mentioned above, this paper makes full use of logging, geology, seismic and other data to carry out fine characterization of fracture with multi-information fusion fracture modeling based on neural network. Firstly, neural network technology is used to predict the fracture density curves of wells with conventional logging data. Secondly, combined the nonlinear fusion of multiple pre-stack seismic attributes including Tectonic Stress Field, P-Wave AVO and Seismic Discontinuity Detection based on neural network, the fracture density probability model is created to predict three-dimensional distribution of fractures. Thirdly, the fracture Intensity model is set up under the constraint of strict variation function analysis and fracture density probability model. Finally, fracture parameters are obtained according to core observation and imaging logging data, and the fracture network is established by geostatistics modeling method and coarsened equivalent into fracture property model. Blind well validation and dynamic validation are used to verify the validity of fracture model. Under the guidance of research results in this paper, the oilfield development plan is optimized and adjusted. It has also been proved that fracture characteristic of new wells is consistent with pre-drilling prediction result. Most areas with relatively high fracture development degree are corresponding to well groups with faster water cut rising rate. Against the backdrop of global oil economic downturn, fracture modeling technology studied in this paper provides strong technical guidance for oilfield development plan, therefore reducing oilfield development risks and maximizing economic benefits.

Real-time location information is essential in the hazardous process and construction areas for safety and emergency management, security, search and rescue, and even productivity tracking. It's also crucial during pandemics such as the COVID-19 pandemic for contact tracing to isolate those who came to the proximity of infected individuals. While global positioning systems (GPS), can address the demand for location awareness in outdoor environments, another accurate location estimation technology for indoor environments where GPS doesn't perform well is required. This paper presents the development and deployment of an end-to-end cost-effective real-time personnel location system suitable for both indoor and outdoor hazardous and safe areas. It leverages on facility wireless communication systems, wearable technologies such as smart helmets and wearable tags, and machine learning. Personnel carries the client device which collects location-related information and sends it to the localization algorithm in the cloud. When the personnel moves, the tracking dashboard shows client location in real-time. The proposed localization algorithm relies on wireless signal fingerprinting and machine learning algorithms to estimate the location. The machine learning algorithm is a mix of clustering and classification that was designed to scale well with bigger target areas and is suitable for cloud deployment. The system was tested in both office and industrial process environments using consumer-grade handphones and intrinsically safe wearable devices. It achieved an average distance error of less than 2 meters in 3D space.

Total organic carbon (TOC) is an important factor for the characterization of unconventional shale resources; which is currently evaluated by either conducting extensive laboratory work, using empirical correlations developed based on linear regression analysis, or applying artificial intelligence (AI) techniques. The AI models approved their efficiency for TOC estimation compared to the use or empirical correlation and they have the advantage of providing a continuous TOC profile compared to the laboratory-based evaluation. This study is aimed to evaluate the predictability of the TOC using two AI models namely functional neural networks (FNN) and support vector machine (SVM). The AI models were trained to estimate the TOC based on well log data of gamma ray, deep resistivity, sonic transit time, and bulk formation density, more than 500 datasets of the well logs and their corresponding core-derived TOC collected from Barnett shale were used to train and optimize the AI models. The predictability of the optimized AI models was then tested on other data from Barnett shale and validated on unseen data from Devonian shale. The ability of the optimized AI models to estimation the TOC for Devonian shale was compared with Wang's density-based correlation (WDC) which was developed recently to estimate the TOC for Devonian formation. The results showed that the AI models predicted the TOC with high accuracy, and they overperformed WDC in estimating the TOC for Devonian formation. For the validation data, FNN model overperformed SVM in estimating TOC with average absolute percentage error (AAPE) of 12.0% and correlation coefficient (R) of 0.88, while SVM model predicted the TOC with AAPE and R of 14.5% and 0.86, respectively, and WDC estimated the TOC with high AAPE of 34.6% and low R of 0.61.

During well testing operations, an incorrect burner combustion may pose a risk of adverse impact on the environment or people. The combustion efficiency is assessed by personnel who observe the flame. This practice lacks consistency and may pose challenges, from environmental and safety considerations to data quantity and quality. An automatic alarm triggered by suboptimal combustion is needed. In this paper, we propose a solution that uses a deep neural network that learns from flame videos to define the quality of the combustion. Flame features that help determine the combustion quality are the flame color, the color of the smoke around the flame and the flame shape and pattern, among many others. To extract the features relevant to the problem and qualify the combustion, we performed supervised learning using a convolutional neural network. We trained the network using videos of flames from burner combustion; the network extracts relevant features and classifies the flame in video images as acceptable or undesirable. We labeled a set of flame videos obtained from well testing operations. The videos include various flame conditions, in day and night scenarios and acceptable and undesirable combustion contexts. We trained the convolutional neural network using the labeled flame videos. The network prediction accuracy is 100% on the training set and 96.8% on the test set. The misclassified images are in the transition periods between acceptable and undesirable combustion events. We demonstrated the potential of a deep learning solution using a convolutional neural network for classifying the combustion quality of burner images as acceptable or undesirable. The results are promising, and they show that this solution is a good candidate for real-time burner efficiency monitoring and automatic alarm triggering and optimization.

Nonconventional wells (NCWs) are applied to increase the well deliverability and access the difficult formations. The nonconventional wells have been used to refer for the advanced wells such as highly deviated, horizontal, fishbones or multilateral wells. These wells offer a great potential to maximize the hydrocarbon recovery, however, it is difficult to predict their performance. In the literature, numerous mathematical models were developed to predict the well-productivity. However, the available models have been developed by employing one or more simplifying assumption(s), which may lead to over or under estimate the hydrocarbon production. This paper presents an effective technique to estimate the productivity for nonconventional wells. In this work, artificial intelligence (AI) technique was utilized to determine the well-productivity for wide range of conditions. The developed models can determine the well performance without introducing the complexity associated with the numerical approaches. Artificial neural network was utilized to estimate the hydrocarbon production for two types of nonconventional wells; fishbone multilateral and hydraulically fractured horizontal wells. Reliable models are presented to quantify the performance of nonconventional wells producing from heterogeneous and anisotropic formations. The developed models evaluate the importance of reservoir properties and well configuration on the well deliverability. Total of 850 data sets were utilized to construct the intelligence models and to validate the prediction performance. Moreover, mathematical equations were extracted utilizing the optimized ANN models. The extracted correlations showed acceptable prediction errors, the absolute error around 7.4% in average. The novelty of this work is that effective models are proposed to quantify the productivity of nonconventional wells. The proposed models can be utilized to refine commercial software to narrow down the deviations between the actual measurements and simulation outputs. Also, this work can contribute to enhance our understanding of the oil-field management by improving the prediction of well deliverability. Consequently, this study can help in optimizing the well planning for complex wells such as hydraulically fractured horizontal and fishbone multilateral wells.

Image logs such as Microresistivity give the interpreter an indication of bedding patterns and depositional environments. This allows him, after correlation with an analysis of selected curve shapes (SP, GR) to interpret the grain size along the well. Both image analysis (from image logs) and data aggregation are areas where Deep Neural Networks have proven efficient, but their accuracy in the field of grain size prediction is still limited, hence not useful yet for an interpreter. Since relevant and complete log data is usually only available on small intervals, networks are overly trained on hand-picked well sections, while ultimately predicting a softmax-activated vector of all grain size "probabilities". Working with data from deepwater wells off the Gulf of Mexico, we have been able to account for uncertainties and scattering in the training data to produce a set of single-tasked, multi-input, multi-output classification CNNs. Combined, these networks produced more accurate results than their multi-category counterparts, with accuracy over 87% on test wells. Using the computed metrics (recall, precision) of these single networks, a confidence curve of the overall algorithm is produced, highlighting high-confidence high-accuracy areas (85% of test wells), as well as areas where human interpretation is required. This approach combines Machine Learning prediction and uncertainty analysis to reshape, using deterministic methods, the "black box" predictions of the Neural Network. This provides a methodology which is more accurate than conventional Machine Learning, while also being aware of its own inaccuracies and highlighting them. It uses complex network architectures and data ingestion mechanisms, designed to generate results that can be exploited by a petrophysicist in charge of log interpretation.

As artificial intelligence continues to revolutionize the industries around with innovations that make work easy and fast, its’ use in the Oil and Gas Industry has been quite limited. With the technology available to us, artificial intelligence ought to revolutionize the Oil and Gas Industry, mainly by time-saving and increasing the ease of operation. This study uses deep learning to predict the lithology of the formations encountered with the use of open-hole log data. It also works to calculate the porosity of the hydrocarbon bearing formation using the density and neutron values. The trained program analyses and assimilates the log values to indicate the lithology, reducing the time of the operation during the exploration stage. If used as a real-time indicator, the program ought to ease the drilling operation. With drilling being one of the costliest operations, any ease in the process that saved time would be a boon to the Industry. The paper gives a summary of the technique used and discusses the work that could further strengthen the operation at hand.

The amount and type of clays in reservoirs have a significant impact on formation evaluation and reservoir performance studies. Currently, clay typing requires either reservoir cores for laboratory analysis or advanced logs such as elemental spectral mineralogy logs, both are available in only a small fraction of wells drilled. In this study, we will explore a possibility of using commonly measured logs to estimate clay volumes. Specifically, the logs used for the study are formation resistivity (RT), total porosity (PHIT) and gamma ray (GR). Since there are no known relationships relating these logs with clay volume, machine learning has been used for data analysis and parameter prediction. This is followed by exploring the possibilities of using array induction resistivity measurements to classify clay types downhole. An important property of clay minerals is their ability to adsorb ions on their exposed surface, which is measured by its cation exchange capacity (CEC). We have developed a method of using induction resistivity data to extract CEC downhole and displayed as depth profiles. There are four major types of clays commonly encountered in the oil fields: Kaolinite, Chlorite, Illite, and Smectite, with their CEC values ranging from low to high. Since each type of clay has its own CEC value, thus a synthetic CEC depth profile for any clay can be constructed if its volume fraction is known. On the other hand, CEC derived from the downhole resistivity data represents the combined effects of all the clay types presented in the formation being surveyed. By comparing the resistivity-based CEC profile with the synthetic ones, it is possible to define a volume fraction for each clay type, for the purpose of clay typing. On the other hand, based on a previous developed method, total CEC representing the combined effects of all clays can be extracted from induction resistivity logs. By comparing the resistivity-based measured total CEC with the synthetic type curves, clay typing from downhole induction resistivity measurement is achieved. A workflow is developed for the application. The proposed methodology was tested on the logs from six wells. The results indicated that RT, PHIT and GR logs have strong correlations with clay volume and the model trained with these logs could be used to predict clay volumes for blind datasets. The workflow for clay typing was tested on the logs from two wells with positive results.

Gas turbines age and sometimes have problems that may lead to a trip or failure event. This paper describes how one can reliably forecast such problems at a future date so that the problem can be dealt with in the present. These methods are data-driven by relying on machine learning to develop the mathematical description for the gas turbine. First, all instrumentation is combined in a state of the turbine that is forecasted into the future using a long short-term memory (LSTM) network. Second, this forecast is analyzed by a neural network trained to recognize normal healthy behavior in order to identify problem. Third, problems can be diagnosed in some special circumstances by recognizing their fingerprint using past examples of the same failure mode; this is particularly relevant to vibration data.