Laboratory experiment is the main tool for determining fluid properties such as saturation pressure, formation volume factor, fluid density and viscosity. However, both sampling and experiments are limited by cost and time. On the other hand, the information about the fluid properties are essential for reservoir engineering works such as reserves and productivity calculation. Therefore, empirical correlations are commonly used as an alternative method. Researchers developed empirical correlations by curve-fitting experimental data. The variation in the correlations is partly due to different datasets. It explains why each correlation more accurately estimates the fluid properties in some cases/regions than in the other cases/regions. In this study, we propose a technique of using surrogate models and the available laboratory database to estimate the fluid properties. Two surrogate models are studied in this paper e.g. universal kriging and neural networks. A comparative analysis is being performed between the proposed technique and known correlations used in the industry. The study shows that the proposed method demonstrates better estimation than the published empirical correlations. This paper has a potential to become a guideline for engineers who would develop an estimation tool with the use of experimental fluid database.

Summary In a recent paper, we published a machine learning method to quantitatively predict reservoir fluid gas/oil ratio (GOR) from advanced mud gas (AMG) data. The significant increase of the model accuracy compared to traditional modeling approaches makes it possible to estimate reservoir fluid GOR based on AMG data while drilling, before the wireline operation. This approach has clear advantages because of early access, low cost, and a continuous reservoir fluid GOR for all reservoir zones. This paper releases further study results to predict other reservoir fluid properties in addition to GOR, which is essential for geo-operations, field development plans, and production optimization. Two approaches were selected to predict other reservoir fluid properties. As illustrated by the reservoir fluid density example, we developed machine learning models for individual reservoir fluid properties for the first approach, similar to the GOR prediction approach in the previous paper. As for the second approach, instead of developing many machine learning models for individual reservoir fluid property, we investigated the essential properties for equation of state (EOS) fluid characterization: C6 and C7+ composition and the molecular weight and density of the C7+ fraction. Once these properties are in place, the entire spectrum of reservoir fluid properties can be calculated with the EOS model. The results of reservoir fluid property prediction are satisfactory with both approaches. The reservoir oil density prediction has a mean average error (MAE) of 0.039 g/cm3. The accuracy is similar to the typical density derived from the pressure gradient from wireline logging data. For the essential fluid properties required for EOS model prediction, the overall accuracy is less than the laboratory measurements but acceptable as the early phase estimations. The reservoir fluid properties predicted from the EOS model are similar to the predictions from individual machine learning models. We applied the field measured AMG data into the reservoir fluid property models and achieved good results, as illustrated by the reservoir fluid density example. The previous paper completed the methodology to predict all reservoir fluid properties based on AMG data. This work paves the way to generate a complete reservoir fluid log for all relevant reservoir fluid properties while drilling. The method has a significant business impact, providing full coverage of reservoir fluid properties along the well path in the early drilling phase. The advantage of providing reservoir fluid properties in all reservoir zones while drilling far outweighs the limitation of somewhat reduced reservoir fluid property accuracy.

In a recent paper (Yang et al., 2019a), we published a machine learning method to quantitatively predict reservoir fluid properties from advanced mud gas (AMG) data. This approach has clear advantages due to early access, low cost, and a continuous reservoir fluid prediction for all reservoir zones. In this paper, we demonstrate how real time reservoir fluid logs are generated and compare the results with PVT samples or production data from the same well. We develop a workflow of generating reservoir fluid logs from AMG data and PVT database. The workflow consists of two main processes; first a quality assessment of AMG data and second the computation of reservoir fluid properties (in this paper we use gas oil ratio). The entire workflow is written in python and embedded into existing commercial petrophysics softwares. The final product of the workflow are three log tracks which we call the reservoir fluid logs and those are 1) the concentration readings of the AMG data, 2) the QC metric score, and 3) the predicted GOR log. These three logs are plotted together with other standard open hole logs such as gamma ray, neutron-density, sonic and resistivity log to get a more comprehensive formation evaluation. Reservoir fluid logs derived from AMG data has two main advantages. First, it is the only approach to acquire continuous reservoir fluid properties along the well path. The continuous fluid profile can be used to understand the variation of reservoir fluids in both vertical and lateral direction. The second advantage is that the reservoir fluid log is obtained while drilling and therefore the information can be used to optimize the drilling process or the downhole sampling program during wireline operation. In this paper, we demonstrate the application of the reservoir fluid logs in four conventional field cases. In the first study case we show the benefit of using the reservoir fluid logs in a horizontal well as a substitute for downhole fluid sampling. In the second case study, we demonstrate how the reservoir fluid log is utilized to optimize the downhole fluid sampling program which results in reducing the subsurface uncertainty. Next, we exhibit the use of the reservoir fluid logs to locate gas oil contact in a case where pressure data does not show clear distinction of gas and oil gradient in the reservoir. In the last example, we illustrate the use of reservoir fluid knowledge from AMG to characterizing the fluid variation across a field. The field applications demonstrate the success of the new method for conventional reservoirs, provided good-quality AMG data are available.

An accurate estimation of viscosity values is imperative for an optimal production and transport design of hydrocarbon fluids. Based on this requirement, cost efficient and reliable empirical correlation models are highly profitable. While there are numerous correlation models from literature, a consistent correlation model is still needed, as most models are inadequate to predict an accurate oil viscosity using unbiased data. This study aims to develop new and improved empirical viscosity correlations through available field measurements on the NCS. The performance of the proposed models is then studied through a comparative analysis with published correlations from literature. I would like to express my sincere gratitude to my supervisor at Statoil ASA, Knut Kristian Meisingset, for giving me the opportunity to write my Master's Thesis for Statoil ASA, his support and interest in my work have been instrumental for the quality of this thesis. Special thanks go to Ibnu Hafidz Arief for his guidance to develop the machine learning algorithms. I am exceedingly grateful for his enthusiasm throughout this semester, the feedback and discussions have been greatly appreciated. Lastly, my appreciation goes to my faculty supervisor, Professor Runar Bøe, for his valuable input, and for letting me pursue my thesis in cooperation with Statoil ASA.

Inter-well tracer test (IWTT) is a method used to track the movement of injection fluid and identify the field connectivity. The mechanism of IWTTs consists of injecting slugs of tracer in the water injector and observing the tracer at the producers from water samples. One of the advantageous of the IWTT is the uniqueness of tracers which indicates to a specific tracer origin. However in fields which apply produced water reinjection, the original location of the tracer might not be determined. This is caused by the reinjection of produced tracer in other injectors. The tracer reinjection could add noises to the tracer data which might lead to misinterpretations. The basic idea to overcome this problem is to minimize the noises so that it cannot be detected. It could be achieved by maintaining the noise level under the detection limit of the analytical tool (gas chromatography/mass spectrometry). The noise created from the tracer reinjection is directly proportional to the amount of tracer being re-injected and the amount of tracer being re-injected is directly proportional to the initial amount of tracer being injected in the first injector. Therefore in order to minimize the noises, the initial tracer amount should also be minimized. However, too little amount of tracer might prevent the tracer from being observed in the target producers. This paper discusses the methodology in optimizing the tracer amount. There are basically 4 criteria that need to be considered in optimizing the tracer amount. First, the tracer amount should be sufficient for the tracer to be detected in the target producers. Second, the tracer amount should be minimal to keep the noises below the detection limit. Third, the tracer amount should also results in minimum delay of tracer reading in the producers. The last criterion is to keep a minimum cost for the IWTT project. This methodology has high potentials to serve as a guideline for reservoir engineers when designing an IWTT in PWRI fields. In conclusion, by having a proper design, the IWTT in PWRI fields can give the same benefits as in the non-PWRI fields.