By International Petroleum Technology Conference (IPTC) Monday, 25 March 0900-1600 hours Instructors: Olivier Dubrule and Lukas Mosser, Imperial College London Deep Learning (DL) is already bringing game-changing applications to the petroleum industry, and this is certainly the beginning of an enduring trend. Many petroleum engineers and geoscientists are interested to know more about DL but are not sure where to start. This one-day course aims to provide this introduction. The first half of the course presents the formalism of Logistic Regression, Neural Networks and Convolutional Neural Networks and some of their applications. Much of the standard terminology used in DL applications is also presented. In the afternoon, the online environment associated with DL is discussed, from Python libraries to software repositories, including useful websites and big datasets. The last part of the course is spent discussing the most promising subsurface applications of DL.
Brunei offshore platforms are home to hundreds of maturing wells in need of ongoing interventions. Offshore operations in Brunei face several obstacles, (i.e., weather conditions, ageing platform facilities, limited lifting capability, and limited workspace), as well as tight work schedules that make the work challenging.
As with other mature fields, the Brunei wells need high efficiency operations to reach production targets. These challenges can be addressed with a purpose built compact semi-submersible vessel (CSS) with dynamic positioning (DP 2) equipped with a full catenary coiled tubing unit, a pumping unit with flowback capability, and a dedicated slickline unit.
Dual hull design with a compensated gangway increases the weather working envelope of the vessel. The coiled tubing catenary system with a reel turntable helps enable coiled tubing unit flexibility during rigup and work under varying weather conditions. Integration of the vessel and the coiled tubing unit helps enable a 24/7, 365 day work unit.
Average downtime caused by weather decreased by up to 10%, averaging 8.5% in 2 years, compared to previous work vessels with an average between 15 and 18% downtime because of weather.
Further efficiency improvement is gained through use of fit for purpose equipment. A 35 ton jacking frame helps enable injector and pressure control equipment stack up to be made up, function, and pressure tested offline. A small footprint flowback package was introduced that reduced the total number of lifts from 12 to 6, saving two hours lifting time per rig up/down. Overall rigging up time was reduced by approximately 20% with the improvements to equipment setup.
The reduced equipment necessary on the platform enabled wireline and coiled tubing to operate concurrently. This enables 24 hour wireline interventions to be executed offline more efficiently. Time savings for intervention completion were approximately 62%. This setup enables more efficient use of existing resources to complete the work scope.
The setup, collaboration, and execution on the vessel demonstrate the opportunity for improvement, which is important under current industry conditions, and help enable a cost effective yet robust operation.
The primary aim of this study was to design a guideline in the form of a flow chart that can potentially be used in practice when optimal sand control methods and implementation of completion techniques are considered in sand producing wells. The flow chart could also serve as a tool during the decision-making process where sand control is deemed necessary. Distinct reservoir data and key parameters were incorporated into the designing of the flow chart. This paper predominantly focused on two of the most common techniques used currently in practice; sand screens and gravel pack.
A retrospective analytical review was carried out. A systematic search was undertaken to analyze sand control methodologies used in recent studies to ascertain key considerations undertaken when choosing gravel pack and/or sand screen. Studies selected were based on a predetermined set of inclusion and exclusion criteria. Recurrent pattern that existed when choice of a specific technique was identified; a list of criteria that was considered when selecting gravel pack and/or sand screen techniques was developed. Information and data obtained were then eventually integrated in stages into designing a concise flowchart.
List of criteria developed when contemplating sand control and completion methods were as follows, 1. Determine rock mechanics, 2. Study individual reservoir conditions, 3. Note lithological changes, 4. Obtain well data, 5. Characterize formation sand, 6. Select gravel size, 7. Select screen and size, 8. Select completion method and 9. Evaluate the potential cost and economical outcome. The development of flow chart then began with the categorisation of key information into four significant stages as follows, sand prediction, sand analysis, sand control and completion method. Looking at the step 1 to 9 of the list of criteria and the four phases of the flow chart together, a more structured and integrated thought process took place. When the sand analysis stage was referred to, criteria 1 to 5 were determined simultaneously. Then, the sand prediction stage was referred to in the flow chart where upon obtaining the sand production rate it could be determined if the sand produced was low/manageable. A choice to live with the sand produced is made if so. If the rate was high, then the rest of criteria 6 to 9 as per the list above were determined and advancement is made to the next stage flow chart according to the sections. Both, the criteria list and flow chart can be used in parallel as guidance when implementing respective techniques in an individual sand-producing well.
Gelinsky, Stephan (Shell International E&P) | Kho, Sze-Fong (Shell International E&P) | Espejo, Irene (Shell International E&P) | Keym, Matthias (Shell Malaysia) | Näth, Jochen (BSP) | Lehner, Beni (BSP) | Setiana, Agus (BSP) | Esquito, Bench (SDB) | Jäger, Günther (SDB)
Prospects below or near shallower producing fields can be economically attractive yet also risky since reservoir presence may be uncertain, reservoir quality can be poor, and high overpressure and temperature can make drilling and logging deeper prospects difficult. Systematic integration of relevant subsurface data from thin section to basin scale allows to seismically identify reservoir presence, and to predict reservoir quality for applicable rock types via burial histories. On an intermediate well log to seismic scale, a predictive rock physics modeling approach links reservoir and seal rock properties to seismic amplitude data to polarize the prospect's geologic ‘probability of success'. Particular challenges in the offshore Brunei study were very fine-grained deposits and non-vertical tectonic stresses associated with compressional settings. Both make porosity predictions that leverage complex burial histories rather than relying on extrapolated depth trends quite challenging - yet the integrated approach remains the best option to identify deep reservoir quality sweetspots that a favorable stress and temperature history may have preserved for certain reservoir rock types in certain locations.
The prolific petroleum system offshore Brunei features two major sediment fairways, the Baram and Champion river systems, and a variety of depositional environments, ranging from high NtG topsets inboard over shallow marine slope settings to deepwater turbidites outboard (
Over recent years, many authors have proposed to compensate the absorption loss effects inside of the imaging process through the use of an attenuation model. This is necessary in the presence of strong attenuation anomalies. Q tomography has been developed for estimating this attenuation model but is generally limited to estimating attenuation in predefined anomaly areas. In this paper, we show how shallow gas pockets are revealed automatically by using a high-resolution volumetric Q tomography on the complex offshore Brunei dataset. A key component of our approach is the estimation of effective attenuation in pre-stack migrated domain through accurate picking of the frequency peak. Estimated Q-model is then used to compensate for absorption in the imaging process.
The Brunei region is considered as a complex area known for its gas escaping features over folded structures, producing shallow strong absorption anomalies. These strong anomalies seriously mask the coherency of the structure beneath.
Typically, the overall effect on the signal is that higher frequencies are dimmed more rapidly as the signal propagates through these very attenuating media. This results in a loss of signal resolution. Conversely, the attenuated signal carries additional information that can be useful in locating such gas pockets.
Measured attenuation can be compensated by applying processes such as the early techniques of inverse-Q filtering (Wang, 2002). More recently, stronger compensation due to gas or mud was included directly in the imaging process (Xie et al., 2009; Fletcher et al., 2012) through an interval Q model computed by tomography (Xin et al., 2008; Cavalca et al., 2011; Xin et al., 2014, Gamar et al., 2015). Generally, effective Q quantities are then inverted to produce a 3D interval Q model. The main purpose of tomography is to de-noise effective Q measurements in a model-consistent manner. Because the tomographic inverse problem is poorly constrained due to a difficult estimation of effective attenuation, a priori information is introduced to guide the inversion.
We present a robust workflow that uses Q tomography for converting dense inhomogeneous prestack effective Q measurements into a 3D model-consistent interval Q. To compute the effective Q volume in the pre-stack domain, we have used the method proposed by Zhang and Ulrych (2002) based on the shift of the frequency peak. Since the frequency peak (frequency at maximum amplitude) is very sensitive to the noise, we increase the signal/noise ratio by using the autocorrelation of the signal rather than the signal itself. This improves the resolution of the frequency peak value and thus the accuracy of effective Q estimation. We apply the workflow on Brunei offshore dataset to localize shallow gas pockets without any a priori information on their positions. This was made possible thanks to an adaptation to Q tomography of non-linear slope tomography (Guillaume et al., 2011) using an accurate effective Q volume picked from pre-stack migrated gathers.
This article, written by Special Publications Editor Adam Wilson, contains highlights of paper SPE 170584, “Dynamic Fault-Seal-Breakdown Investigation - A Study of Egret Field in the North Sea,” by P.P. Obeahon, SPE, G. Ypma, and O.U. Onyeagoro, SPE, Shell, and A.C. Gringarten, SPE, Imperial College London, prepared for the 2014 SPE Annual Technical Conference and Exhibition, Amsterdam, 27–29 October. The paper has not been peer reviewed.
The ability to predict the effect of faults on locating remaining hydrocarbon is critical to optimal well-placement, reservoir-management, and field development decisions. The tools and techniques available for realistic differentiation between sealing and nonsealing faults have presented a great challenge to the industry. This paper discusses the results of an integrated study that incorporated detailed geology and reservoir engineering to understand production behavior of a complexly faulted high-pressure/high-temperature field in the North Sea.
Predicting fault-seal breakdown is a challenging task because it involves many interrelated factors and complex relationships. Knowledge of these factors is both nonunique and subjective. Most faulting processes have been studied in isolation, and the relationships among many of the processes are understood poorly.
Reservoir depletion can, in principle, induce stress paths capable of reactivating intrareservoir faults and, hence, potentially cause breakdown of their sealing integrity. Fault-seal breakdown may also be invoked falsely where oil/ water contacts change across a fault (i.e., the fault is a capillary seal) but the fault does not compartmentalize pressures in production. This apparent seal failure can arise because of pressure communication in the water leg below the oil column. It is not clear why pressure depletion should cause capillary-seal failure. However, publications exist that attempt to attribute production behavior observed in fields to fault-seal breakdown in a production realm, because of pressure depletion on one side of a fault.
The first attempts to incorporate geologically reasonable fault properties into production-simulation models involved the calculation of transmissibility multiplier on the basis of absolute permeability and thickness of fault rocks. These calculations do not capture the multiphase behaviors of fault rocks. A key problem with this approach is that a huge number of pseudofunctions needs to be calculated to take into account the large variation in fault properties (e.g., thickness, absolute permeability) and flow rates and whether the fault is going through drainage or imbibition during production. The second attempt involved calculating transmissibility multipliers (also known as seal factors) on the basis of fault permeability. The key problem with this approach is that fault permeability depends on shale gouge ratio and fault displacement alone. The calculation does not capture the impact of reservoir permeability on fault permeability.
Intelligent Reservoir Management and Monitoring has played a key role in the pursuit of improving the hydrocarbon recovery and reducing the development expenditure in the challenging multi-stacked compartmentalized fields which have proved to be perplexing in a number of ways which include preventing or delaying water breakthrough, extenuating wellbore instability, sand production etc. Reservoir-management and monitoring options have been greatly improved in recent decade by smart completions comprising of downhole monitoring and control equipments like permanent down-hole gauges to have "eyes" into the reservoir and to monitor performance for each zone; dynamic active flow control valves, which aid in equalization of the reservoir inflow into the wellbore; and the SCADA system which enables the real time monitoring and control of the downhole and surface equipment remotely from the control room.
The Maharaja Lela Jamalulalam field (MLJ), located in Brunei Block B, was discovered in february1990 and has been producing oil, gas and condensates since 1999. It is operated by Total E&P Borneo B.V. (TEPB) with a 37.5% interest. The remaining interests in the Block B Joint Venture (BBJV) are held by Shell Deepwater Borneo Limited (SDBL), 35% and Petroleum Brunei ExPro (PBE), 27.5%. Figure 2 MLJ installations layout schematic and Lumut plant Block B is located approximately 50 km offshore Brunei in shallow waters (65m) and comprises the main sub-block (B4) and 2 separated smaller sub-blocks (B6 and B2) located respectively to the South and East. The main subblock contains the Maharaja Lela Jamalulalam field (MLJ) and the production facilities.
Crampin, Tom (Brunei Shell Petroleum Co. Sdn. Bhd.) | Gligorijevic, Aleksandar (Geoservices) | Clarke, Ed (Shell) | Burgess, Jamie (Brunei Shell Petroleum Co. Sdn. Bhd.) | Chung, Shao-Jung (Brunei Shell Petroleum)
Downhole determination of hydrocarbon phase is a significant subsurface challenge in many highly depleted fields. Reservoir production results in fluid compositional changes and variable hydrocarbon saturation distributions. Standard petrophysical techniques such as analysis of density and neutron porosity logs can give misleading results under such conditions. Most commonly, oil reservoirs can display a neutron-density response indicative of gas. There is significant business impact in error of hydrocarbon phase determination. Mistakes can lead to poor completion decisions, incorrect reserves estimation and suboptimal well and reservoir management.
The fluid phase uncertainty resulting from interpretation of standard Logging While Drilling (LWD) datasets can be unacceptably high. Additional tools or techniques are therefore required. Downhole fluid sampling is one such technique. It is routinely and successfully acquired in exploration and appraisal wells and gives robust fluid phase determination. However, it is not economically feasible for frequent acquisition for in-fill production wells where low cost LWD acquisition is the norm. In addition, overbalanced wells drilled through highly depleted reservoirs lead to acquisition risk in stationary openhole logging techniques. Advanced Mud Gas logging (AMG) is an established tool for delivering real-time quantitative fluid composition in exploration, appraisal and early production wells. However, successful applications in highly depleted fields have not been published as AMG analysis can be complicated by compositional changes. In this paper we present a case study calibration of AMG with downhole fluid samples resulting in a robust, cost effective and safe tool for improved hydrocarbon phase determination in depleted reservoirs.
Many techniques are used to determine hydrocarbon phase but all of them can be impacted by production related changes to reservoir fluids. The neutron-density "cross-over?? is the most common gas identification tool (Figure 1). It results from an anomalously low neutron porosity reading in gas, due to low hydrogen index (HI), and an anomalously high density porosity reading, due to low fluid density. A second traditional technique is the neutron near count to far count ratio. The near detector reads largely in the near wellbore invaded zone where high mud filtrate saturation results in a high HI and a relatively low count rate when overlain with the far detector, which reads deeper into the formation, past the invaded zone, resulting in a relatively high count rate if gas is present.