Cement evaluation data acquired in oil and gas wells for confirmation of zonal isolation, channeling in cement behind casing and well integrity. All available technologies for cement evaluations are primarily measurements of acoustic parameters like amplitude of first arrival, full waveform recording of refracted wave, impedance and attenuation or there a combination. Generally, operationsPetrophysicist, petroleum engineer or service providers are responsible for evaluation of cement bond logs and propose remedial jobs if required. This cement bond log interpretation is quite subjective if performed through pure visual interpretation and its accuracy depends on objective of well, work pressure and experience of interpreter. This paper talks about how ADNOC Onshore has leveraged machine learning (ML) for interpretation of various cement bond logs from several service providers. In green fields, cement evaluation is very important in all new wells to ensure good cement quality and zonal isolation and it is also equally important in in brown fields where 1000s of wells were drilled where well integrity issues are of common occurrence. Major challenges in evaluation are inconsistency and human bias in interpretation and as a result, interpretation may vary from one interpreter to the other. Authors have tested different ML techniques (Random Forest Classification & Neural Net) and smart way of data training. Final recommendation is to use nested models instead of single model. In this technique, input measurements and required solutions will be classified and divided into different classes, a separate ML model will be built for each class and combine all the models to get final cement evaluation and recommendations. Final results include data quality flags, cement bond quality, zonal isolation, channeling, micro annuals, areal cement map (if analyzed field wise), anomaly maps (if analyzed data from offset wells at different times), and recommendation. In this project, a simple easy to use user interface has been developed to browse the cement logs and use the trained ML models to predict the cement evaluations with a click of a button. This ML based cement bond evaluation is proved to be very effective and saved 75% of efforts by operational Petrophysicists. The interpretation accuracy has been significantly improved. This method has potential to be used at rig site. These models are very cost effective with, minimum human bias, and improves consistency as well as independent of wells or type of reservoirs. These models are tested across ADNOC Onshore Wells and can be extended to any well irrespective of geographic locations. This paper discusses machine-learning approach, evaluation of various algorithms and testing results for cement evaluation log from various service providers.

Modeling of water saturation in carbonate reservoirs is a challenging task which becomes more difficult if the reservoir is highly affected by diagenesis processes. This paper addresses a robust water saturation modeling workflow that encompass multiple approaches followed to handle an onshore Abu Dhabi carbonate reservoir that have complex pore network. This covers a holistic reservoir petrophysical evaluation all the way to water saturation modeling for dynamic simulation studies. The studied reservoir was deposited in high energy shallow, water platform environment prevailing during Early Cretaceous time. It has a large areal extent with low average porosity and moderate to low average permeability. Considering the impact of porosity exponent (m) and water resistivity (Rw) on the water saturation evaluation, the workflow was adjusted to integrate all available log, core and well test data. By integrating the formation resistivity factor (FRF) at different net confining stress and petrography data, relationship for variable porosity exponent (m) versus porosity was developed. The salinity variation was observed across the field (east to west) from the well test and Pickett plot. In water saturation estimation, salinity taken from one side of the field (east/west) causes either under or overestimation of water saturation by 18% on the other side of the field which has a Considerable impact on the STOIIP. To address this problem, variable formation water resistivity (Rw) was generated from log data. This was validated using Porous Plate capillary pressure data measured using samples collected from different areas of the field. The destructive dolomitization at the bottom part of the reservoir resulted in the large transition zone and sporadic occurrence of poor lithofacies poses a great challenge in developing a saturation height function. The current case study also evaluates the best possible saturation height function which can explain the variability of water saturation across the field.

Carbonates exhibits diverse flow characteristics at pore scale. Petrographic study reveals micro-level heterogeneities. Thin sections are key to assess reservoir quality although these are images and interpretations in text format. Thin section microscopic analysis is descriptive and subjective. To an extent, optical point counting is routinely used quantitatively to estimate porosity, cement, and granular features. Overall, thin section descriptions require specialist human skill and an extensive effort, as it is repetitive and time consuming. Thus, a manual process limits the overall progress of rock quality assessment. There is no recognized method to handle thin sections for direct input with conventional core data due to its image and descriptive nature of data. An automated image processing is one of the emerging concepts designed in this paper to batch process thin sections for digital reservoir descriptions and cross correlating the results with conventional core analysis data. Thin section images are photomicrographs under plane polarized light. Initially, denoise and image enhancement techniques were implemented to preserve elemental boundaries. Computational algorithms mainly, multilevel thresholding and pixel intensity clustering algorithms were programmed to segment images for extracting elements from segmented regions. The extracted elements were compared with original image for labeling. The labeled elements are interpreted for geological elements such as matrix, pores, cement, and other granular content. The interpreted geological elements are then measured for their physical properties like area, equivalent diameter, perimeter, solidity, eccentricity, and entropy. 2D-Porosity, polymodal pore size distribution, mean pore size, cement and granular contents were then derived for each thin section image. The estimated properties were compared with conventional core after calibrating with laboratory NMR data. The whole process is automated in a batch process for a specific reservoir type and computational cost is analyzed for optimization. 2D-porosity is in excellent agreement with core porosity, thus reducing uncertainty that arises from visual estimations. Scale related issues were highlighted between 2D porosity and core porosity for some samples. Polymodal pore size distributions are in good correlation with NMR T2 distribution compared to MICP distributions. The correlation coefficient was understood to be equivalent to surface relaxivity. A digital dataset consisting of 2D porosity, eccentricity, entropy, mean pore size, cement and grain contents is automatically extracted in csv format. The digital dataset, which was previously in text format in conventional analysis, is now a rich quantitative dataset. This paper demonstrated a unique and customized solution to extract digital reservoir descriptions for geoscience applications. This significantly reduced the subjectivity in visual descriptions. The solution presented is scalable to large number of samples with significant reduction in turnaround and effort compared to conventional techniques. Additional merit is that the result from this method has direct correlation to conventional core data for improving rock typing workflows. This paper presents a novel means to use thin section images directly in digital format in geoscience applications.

The objective of this paper is to present a unique Petrophysical Grouping (PG) approach in a carbonate reservoir located in transition zone. It is very challenging, especially in Carbonate reservoirs, exist in transition zone, to establish PG definitions due to the complexities result from reservoir heterogeneities and diagenesis. Consequently establishing a suitable Saturation Height Function to match the Log derived Saturation is another challenge. In addition, the limited coverage of Mercury Injection based Capillary Pressure data (MICP) as compared to Routine Core analysis (RCA) data provides difficulties in establishing appropriate PG definition. In first step the MICP data was used together with porosity/permeability to define distinctive groups. The PGs were further up-scaled using deterministic and Neural Network (NN) approaches. The best method was chosen by performing a test that compares the Washburn Pore Throat Radius (PTR) with the predicted PTR. To estimate a most representative log based permeability model, independent of water saturation a NN and Self-Organization Map methodologies were adapted. The limitations of MICP samples were handled by using an analog of a larger field with 100s of MICP samples. This was used to propagate the PGs to log domain by utilizing the permeability model. Five PGs were defined using deterministic approach in which the best one is characterized having low displacement pressure, low irreducible water saturation, high pore throat radius and high porosity and permeability responses. Winland was shortlisted after testing other methods as the most applicable PG method in the reservoir as it provides the best correlation with lab PTR (94%) and the shape of WR35, consequently provides good match with computed Sw log and the shape of the PRT curve (Gunter, 2017). Regardless of the good response of NN approach the method was not chosen due to limitation of MICP data. A good relationship of Winland based PGs were found with geology and associated facies indicate strong affinity with the depositional environment and diagenetic overprints on each existing facie associations, hence a permeability model is depicted with confidence. The permeability model was executed for two geographic sectors using density, neutron porosity and GR as main inputs. SOM and NN Permeability were blind tested which resulted in more than 80% match. The predicted Sw matches with log based Sw over the entire field thus the PGs definition and propagation to log domain are considered valid.

Reservoir characterisation in laminated sand shale reservoir has always been a very challenging task. The presence of conductive shales between the resistive hydrocarbon bearing sand reduces its resistivity drastically. It happens as the current flows through them as resistors in parallel. Therefore, the classical shaly sand interpretation grossly underestimates the hydrocarbon potential of laminated reservoir. Over the years many techniques have been developed to solve this problem. The development of latest resistivity tools like triaxial resistivity has also helped in addressing this problem. The introduction of these tools has helped upto a great extent to unlock the reservoir potential. However, the technology is very costly and sometimes may not be possible to run due the limited extent of the reservoirs. In those cases, it is important to find some alternate approach to addres the problem of low resistivity by using conventional logging tools. The current study deals with the case history from the East coast deep water field in India. The field has many drilled wells where the latest resistivity tools have been run. Petrophysical interpretations are done using this resistivity anisotropy data and are available for validation with the new technique. The latest approach uses the conventional resistivity data. The critical part in the study is the type and distribution of the shale within the reservoir. If we can find out the same then using this information true resistivity can be inverted. The approach has been tried on several wells across the different fields and has given good results. The results from the resistivity anisotropy data and the new technique are comparable. Thus a very cost effective method of predicting true resistivity has been developed, which ultimately gives realistic hydrocarbon saturation in the laminated sand shale sequences.