The text below is a brief description of the project topics supervised by professor Lars Imsland. Contact lars.imsland@itk.ntnu.no for further information. Most topics can be continued in a master thesis project if the candidate and supervisor(s) so wishes.

1. Dynamic positioning systems in arctic environments
   1-2 assignments are available in connection with a research project on development of technology for dynamic positioning in arctic environments. Possible tasks are:
   • Use of Unmanned Aerial Vehicles (UAVs) for measuring ice properties in the arctics
   • Observer design for ice motion
   • Online observability/identifiability analysis and regularization in moving horizon observers
2. Model invalidation and parameter/state estimation using qualitative and quantitative data
   Investigate recently developed algorithms for model invalidation, and apply them to pressure control in drilling, for fault identification. This task suits the mathematically inclined student.
3. Control with discrete output (Statoil)
   Co-supervisors: John-Morten Godhavn (Statoil), Einar Skavland Idsø (Statoil)
   Some actuators have discrete outputs, e.g. step chokes. Statoil has experienced some challenges when these chokes have been used to control the pressure in the wells. Step chokes have a resolution of typically 0.5 – 1.0%. The task is to design a nonlinear controller with improved properties compared to a PI controller. It is not desirable with too much variations neither in the choke position nor the process variable. Dymola or Matlab shall be used.
4. Drilling mud property estimator/observer (Statoil/IRIS)
   Co-supervisors: John-Morten Godhavn (Statoil), Gerhard Nygaard (IRIS)
   1. Study mud properties: density, viscousity, etc
   2. Discuss instrumentation and measurements available in drilling (pressure (absolute and differential), flow, temperature, wired pipe, lab, etc)
   3. Run a drilling simulator (WeMod/IrisDrillforMatlab from IRIS) with different mud properties
   4. Design observers to estimate mud properties from available instrumentation. Assume that the real/true properties are available from lab measurements for calibration.
5. 1-2 assignments related to implementation and tuning of basic control layers (ABB)
   Co-supervisor: Olav Slupphaug (ABB)
   1-2 assignments related to implementation and tuning of basic control layers. Concepts may be evaluated on real structures/data from ABB. Further details will be provided.
6. Techniques for efficient covariance propagation in the Extended Kalman Filter based on model reduction
   Even though the (recursive) EKF is considered an efficient algorithm for state estimation, it can still be computationally expensive for large/complex systems. The bulk of the complexity stems from the propagation of the state covariance matrix. In this task, the student will consider using techniques from model reduction to make the covariance propagation more efficient, while the full model still is used for state propagation.
7. Data-reconciliation of ionic analysis data applied on the Vega field
   Co-supervisors: Geir Arne Evjen, Glenn-Ole Kaasa (Statoil)
   Background/Motivation: : The Vega field is a subsea gas and condensate field located in the northern part of the Northsea, approximately 80 km outside Florø. The field consist of the reservoirs, Vega Nord, Vega Sentral and Vega Sør, and started its production October 2010. The well stream is transported 29 km from the reservoir to the Gjøa (operated by Gas de France) platform for further processing . The well stream contains some water in the gas phase and due to cooling of the fluid along the production pipeline, water condenses out from the gaseous phase and forms a water phase. MEG – Mono-Ethylene-Glycol is therefore continuously injected into the multiphase fluid to prevent gas-hydrates from forming in the production pipeline. At the Gjøa platform, MEG/water phase (rich MEG) is separated from the multiphase stream and regenerated to 90 weight% MEG (lean MEG) before it is again, re-injected into the multiphase fluid.
   The gas from the reservoirs contains CO2 which in contact with water, lowers the pH value of the water/MEG phase and increases the corrosion potential of the production pipeline. To reduce the corrosion potential, chemicals (pH-stabiliser and film forming corrosion inhibitor) are mixed into the MEG before it is injected in the production pipeline.
   It is expected that the reservoirs will at some time in the production period, start to produce formation water. The formation water contains several ionic species which in combination with the injected pH-stabiliser, form solid salts (scale precipitation). On the Vega field, formation of scale is prevented by injecting scale inhibitor into the lean MEG, forcing the ionic species from the formation water and from the pH-stabiliser, to remain in the aqueous solution as super saturated ions. However, the desalination unit on Gjøa, does not handle formation water rate above 10 m3/d, and can cause production stop if exceed. It is therefore essential that the formation water rate is monitored properly to prevent production stop.
   The water-MEG phase that is separated from the gas/condensate at Gjøa platform, is analysed on a regular basis and the measurements (all ions, density and MEG weight %) are stored into a database together will relevant online measurements from the Vega subsea field. MEGSim is an in-house application that utilises the measurements into a simple stationary process model, calculating among others, the total formation water rate from the Vega subsea field, such that the handling capacity of the desalination unit is not exceeded. The main principle behind the application is to focus on the mass balance on the Vega field, using known rates and concentrations to estimate e.g. the unknown total formation water in bilinear data reconciliation. The utilised method requires that all concentrations of the source streams are known in advance (together with their measurement uncertainties).
   The process model of MEGSim and the data reconciliation method therein have difficulties to handle large variations in the measured rich-MEG analysis data. In addition, the process model is purely stationary, and hold-ups both in the production pipeline and lean-MEG injection pipeline, are not accounted for; consequently restricting the model to be used in production periods when the injection rate of lean-MEG and gas production rate is stable. In addition, there have been some experience with carry-over, i.e. the lean-MEG contains some ionic species, typically originating from pH-stabiliser, completion fluids, etc, that cause the process model of MEGSim to fail.
   Outline of work: Based on the above, the following tasks will be suitable for a student, where the main goal is to improve the robustness of the process model in MEGSim:
   • Understand the actual subsea process at Vega
   • Familiarise with the process model implemented in MEGSim for Vega and its tuning tool
   • Understand the bilinear data-reconciliation method used
   • Identify when and where the process model fails
   • Study the measured analysis data from the Vega field, with particular focus on the following production periods
     • Start-up of wells
     • Carry over periods
     • Transient periods
   • Identify how uncertainties in specifications influences the reconciled rates of the process model
   • Retune process model based on experience from measurements and effect of specifications.
   • Identify possible model adjustments. Implement and retune model.
   • Model reduction.