Multidisciplinary approach proves critical
John R. Berry
Niall J. McCormack
Eamonn F. Doyle
The combination of 1,500-m water depth, appalling winter weather conditions, and a recent geological history that greatly increased the uncertainty concerning stratigraphy, pore pressure, fracture, and overburden gradients presented a major challenge to planning and drilling of a deepwater exploration well offshore Norway.
In the winter of 2001-2002, BP drilled the Havsule exploration well in 1,495 m of water offshore mid-Norway using a dynamically positioned semisubmersible. The objective was to test several potentially hydrocarbon-bearing seismic features in the Tertiary and Upper Cretaceous, and gather sufficient data to enable complete verification of the prospect. The well was planned and drilled as a rank wildcat.
Pore pressure predictions from seismic and basin modeling were merged for final prognosis.
Because of the uncertainties inherent both generally in deepwater exploration and specifically in this well, BP assembled a team of formation-pressure and wellbore-stability specialists that was involved in all phases of the well operation from the start of planning, throughout the drilling of the well and in the post-well evaluation.
The well was originally planned to be spudded and completed in the spring and early summer of 2001 to maximize chances of benign weather conditions in this extremely exposed area of the Norwegian shelf. Repeated delays in rig availability, however, resulted in the well being drilled between December 2001 and February 2002, during one of the worst weather periods in recent years, which seriously affected the drilling time curve. It also meant that the entire planning process had to be restarted to take account of a completely different set of circumstances.
The determination of a reliable pore pressure and fracture gradient profile is of great importance at the planning stage of any well. In this stage, studies of seismic interval velocities and 2D basin modeling were undertaken to assess the likely range of pore pressures encountered. These ranges were calibrated with offset well data and then merged to provide the best estimate and worst case planning for the well.
Determining pore pressures from seismic interval velocities is based on the assumption that there is a consistent regional relationship between acoustic velocity and effective overburden stress. The pore pressure is then determined by applying Tezarghi's effective stress law. Offset well data are used to assess the reliability of the link between shale acoustic velocities and pressure. It is then assumed that shales are sufficiently dominant, and their properties sufficiently uniform, that seismic interval velocities closely align to the shale velocities.
The process adopted was to use offset well data to establish appropriate density and overburden trends. Sonic log velocities/travel times were used to predict pressures in the offset wells using standard model settings and to compare with direct measurements of pressure (repeat formation testers, kicks) and indirect indicators of pressure, e.g., mud weights used and hole conditions experienced. Then standard model compaction trend settings were adjusted where necessary to get the most consistent match between predicted pressure for shale acoustic velocity and actual measured/inferred pressure from the well.
Seismic interval velocities were extracted at the proposed well location. The overburden and adjusted model compaction trend settings were then used to calculate the pressure at the prospect location. A similar process was used to match the fracture gradient of the offset wells to those predicted by the model and apply those trends to the prospect location.
An important consideration in this work was to try to account for the effects of major regional events of recent geological time. If formations have been subjected to greater overburden loading than that occurring currently, the result will be over-compaction and hence the normally pressured compaction trend cannot be predicted from the present day loading. Pressure evaluation software needs to be able to take this into account.
The two relevant "events" are the Storegga slide, an enormous submarine avalanche that occurred about 8,000 years ago in which as much as 600 m of sediment were removed from the surface, and the "rebound" caused by removal of ice sheets over this region in the past 20,000 years, which also adds to the complexity of assessing compaction and under-compaction scenarios.
Various uplift scenarios were tested on the offset well data and results included in the prognosis assumptions.
Fracture gradient estimation
A number of generic fracture gradient predictors were used in an attempt to best match the leak-off test (LOT) data from the offset wells. The result was a simple fracture gradient prediction based on assuming a constant Poisson's ratio of 0.41. This gave a predicted fracture gradient that is slightly lower, by about 0.02 sg, than the LOT values at shallow depth, but slightly higher, again by about 0.02 sg, at greater depths.
Basin modeling for pore pressure prediction was performed in addition to the prediction from seismic interval velocities and is based on the assumption that pore pressure is determined by the rate of escape of water from the sediment pile during burial. Thus, the primary controlling factors on pore pressure are:
- Permeabilities of mudstones, which dictate the rate of vertical flow of water in the system
- Lateral connectivity of high-permeability units in three dimensions, controlling the rate of lateral water flow in the system
- Rate of burial, which, combined with the above, dictates the volume of water in the system.
The most likely geological scenario supplied by the subsurface team was used for the first model build. This defined the permeability structures in the Havsule area and the adjacent Ormen Lange area. The lateral extent of the sandstones in the system was modeled to estimate the maximum and minimum up-dip pressure transmission from the basin centers. The 3D aspect of water flow in sandy units is approximated in 2D sections and the effect of lateral connectivity and the extent of the more permeable units are assessed using multiple scenarios.
Differences between basin model-based and velocity-based predictions led to a compromise in which the final prognosis contained a weighting toward the velocity predictions in the very shallow section and joint weighting in the lower Tertiary and toward the basin modeling predictions in the Cretaceous. This reflects the opinion that velocity predictions can be more reliable at shallow depths, while basin modelling is more reliable at greater depth. The final predicted range of pore pressures included a low-, a mid-, and a high-pressure estimate. A similar series of fracture gradient predictions was produced.
A data acquisition (DA) team was created to look at all aspects of data acquisition during drilling to ensure that all needs were met and eventualities planned for. At least one senior person from each relevant service provider was included in the DA team, and status meetings were held each week. Each service provider had responsibility for following up and reporting on relevant items. This increased efficiency and ability to flag potential problems at the appropriate time, and the DA team was an integral part of the well planning team. This ensured that all interests were correctly addressed at all stages of the well.
Pore pressure planning
Knowledge Systems Inc. (KSI) performed pore pressure planning, monitoring, and evaluation in real time to ensure drilling was carried out as safely and efficiently as possible from start to finish and to allow for pro-active management of the drilling process. This involved the use of Knowledge Systems' Drillworks/Predict software and an experienced operator on board the rig from seabed through to total depth at 3,650 m.
The service used real-time input of data from logging-while-drilling (LWD) and surface data logging (SDL) to calculate automatically pore pressure and fracture gradient and to display all input data and results in real time, giving the client ample warning of pore pressure (PP) changes and providing as accurate as possible an estimate of PP at bit at all times. In addition to continual quantitative PP calculation, all other pressure indicators such as gas, hole conditions, and temperature were monitored and used in the process so the user had access to all relevant information on a single display.
The LWD data imported were gamma ray, resistivity, effective circulating density, and annulus temperature. The SDL data were ROP, WOB, RPM, torque, mud density in, mudflow in, and total gas.
During the planning process, an independent pre-drill study was made in which all available relevant data and information for the closest reference wells were collected and imported into the software to create a regional model that was used as the starting point for the well site monitoring.
A summary of the offset wells included in the pre-drill study and the data types were collected wherever possible for those wells.
All data were imported into, and analyzed in, the software to create the pre-drill model. The model obtained was then used as the start point for PP monitoring while drilling and updated with data in real time.
The operator's PP prognosis was used as input to the drilling program, and the well casing and mud programs were based on this. All pressure prognoses were combined in the Drillworks software for a complete overview of the various scenarios. As drilling progressed, each prognosis was updated and discussed with the team.
The well site PP monitoring service was responsible for real-time import and analysis of all data, real-time update of software model, reporting of PP/fracture gradient (FG), and transmitting PP-mud weight-FG-overburden gradient (OBG) data as required. Additionally, the software model was updated with wireline and vertical seismic profile data as available.
The geopressure monitoring service was mobilized so that monitoring could begin from the start of drilling the 12 1/4-in. pilot hole from seabed.
The service was located in the well site geologist's office on the rig and data imported in real-time via an RS-232 cable, which had been pre-installed by the SDL provider, to provide a direct wellsite information transfer specification (WITS) data transfer link between the Knowledge Systems computer and the SDL system.
The formats for transmission of both LWD and SDL data had been agreed beforehand. The LWD data imported were gamma ray, resistivity, effective circulating density, and annulus temperature. The SDL data were rate of penetration (ROP), weight on bit (WOB), revolutions per minute (RPM), torque, mud density in, mudflow in, and total gas. Thus a drilling exponent (Dxc) could be calculated and corrected using actual measured equivalent circulating density. These transmission formats are completely flexible to handle whatever data are available.
The OBG is a very important input parameter to most quantitative pressure estimation methods, and a lot of effort was made to ensure that the OBG used during drilling was the best possible, using offset density data and location seismic data. At the end of the well, wireline density data was used to calculate a final OBG.
The pre-drill studies had indicated that the exponent method could be used reliably in this area for estimating PP, applying the standard exponents of 1.2 for resistivity and Dxc, and 3.0 for sonic.
A summary of the offset wells included in the pre-drill study. The data types were collected wherever possible for those wells.
The PP estimates were updated constantly as the Dxc and resistivity data were received. The LWD gamma ray was used for definition of shale intervals, using an operator-maintained shale baseline, which was then applied automatically to the porosity-indicating logs to create "shale porosity" logs. Normal compaction trend lines were applied to these shale porosity data sets and used for direct calculation of pore pressure.
The wireline sonic was used in the same way as it became available to provide a separate estimation of PP. The sonic log is normally considered the most reliable for PP calculation purposes and therefore the final PP at the end of each hole section was revised as necessary based on the sonic data. Of course where LWD sonic is available this becomes the most important realtime source of quantitative PP.
Although the calculations of PP and FG were performed automatically, full operator control was maintained so that the results could be tuned interactively according to all relevant data inputs, including hole response, gas and lithology. In fact, the few connection gases, which appeared during drilling, were very useful in corroborating the estimated PP.
Throughout the well the two main real-time porosity indicators, Dxc and resistivity, produced quite similar PP profiles, and when wireline sonic data were added the resultant PP from sonic agreed very well with the PP curves from Dxc and resistivity.
In well planning, it is critical to adopt a multidisciplinary approach to produce the most efficient well design to compensate for subsurface uncertainties, a key factor being pressure conditions. The pre-drill pressure prediction for the Havsule deepwater well was created from an integrated interpretation of offset well data, seismic velocity information, and basin modeling to produce the best possible prediction. During the drilling of the well, real time pressure monitoring by a well site pressure specialist formed an essential part of the process.
To revise the pressure interpretation in real time, the offshore team and the onshore experts who created the original prediction integrated all available data from the well and the pre-drill prediction. By using this integrated approach, it was possible to deliver a clear interpretation of the well pressure at any juncture to make well design decisions and improve drilling success.
SPE/IADC Drilling Conference Paper 79848. The authors would like to thank Havsule license holders, BP, Petoro, ConocoPhillips, and Total for permission to write this article, and BP and Knowledge Systems for providing the resources and information needed. We would specifically like to thank Alan Ford, Knut Hansch, Sam Johnson, Andy Johnston, and Mike McClean.
Osborne, M.J. and Swarbrick, R.E. 1997, "Mechanisms for Generating Overpressure in Sedimentary Basins: A Reevaluation": AAPG Bulletin, v. 81, p. 1023-1041.
Beicap-Franlab Petroleum Consultants, Temispack software for basin analysis.
Knowledge Systems, Drillworks software for geopressure analysis.