Closed-loop reservoir management
Overview
Our research into closed-loop reservoir management is focused on the development of concepts and algorithms to improve hydrocarbon production through the use of systems and control theory. Within the Petroleum Engineering section of the Department of Geoscience and Engineering we particularly address reservoir management aspects, i.e. processes on time scales from months to many years. Our main sources of inspiration are systems and control theory as applied in the process industry, and data assimilation techniques as used in areas outside the E&P industry such as in meteorology or oceanography. Therefore we actively cooperate with research groups outside our Department, such as the Delft Institute of Applied Mathematics (DIAM) and the Control Systems group (Dept. of Electrical Eng.) of the Eindhoven University of Technology. We co-initiated the Virtual Asset Learning and Understanding (VALUE), Integrated System Approach to Petroleum Production (ISAPP, phase I and II) and Recovery Factory programs. The first phase of ISAPP (ISAPP-1) started in 2004 as a joint research project between Delft University of Technology, Shell International E&P and the Dutch applied research organization TNO. Formally ended in 2009, ISAPP-1 involved cooperation of more than 25 PhD students and a similar number of TNO staff on topics in two major themes: Subsurface characterization and flow (SCF) and Production Systems (PS). Upon ISAPP-1 approaching its completion, TNO and TU Delft established a successor project, ISAPP-2. With ENI, Statoil and Petrobras as participants the aim of ISAPP-2 remained the application of systems and control theory to significantly increase hydrocarbon production and recovery. ISAPP-2 was strongly focused on the development of computational tools and the application of concepts to field development plans and real field cases. Moreover, early 2010 Shell and TU Delft established the Recovery Factory program aimed at an extension of the closed-loop reservoir management concept to enhanced oil recovery (EOR). With a total of eight PhD students starting over a two-year time-period, the program addressed a range of topics, covering optimization, data assimilation and forward modelling, as well as experimental research into more fundamental EOR aspects. Both the ISAPP II and Recovery Factory programs have now ended. Over the past 18 years, a large number of Smart-Wells/Fields-related journal publications, conference publications, two books, a book chapter and 30 PhD Smart-Wells/Fields related PhD theses have been completed in our Petroleum Engineering group and the affiliated groups at TU Delft and TU Eindhoven.
Model-based reservoir management
The picture below depicts reservoir management as a model-based controlled process. The system, at the top of the figure, comprises of reservoirs, wells and facilities. Many elements of the system, in particular the subsurface parts, are poorly defined, and we therefore typically use multiple system models that each contain many uncertain parameters to represent e.g. porosities, permeabilities, aquifer strength or initial fluid contact positions. Moreover, the state of the system (i.e. the pressures and saturations in the reservoirs, the pressures and phase rates in the wells, etc.) is only known to a very limited extent from the measured output of various sensors at surface or downhole, and from more indirect measurements such as time-lapse seismics. Also the input to the system is only known to a limited extent (i.e. water injection rates or gas lift rates may be roughly known, but aquifer support may be a major unknown). Nevertheless, system models can be used to optimize the reservoir management strategy through forward simulation. This is typically what is done today with dynamic reservoir models, which are used to simulate the effect of different potential development scenarios under different sets of assumptions. In terms of measurement and control theory this is an 'open-loop' control strategy.
Program elements
Data assimilation
Because of geological uncertainties, reservoir models are usually only a very crude approximation of reality, and therefore their predictive value is limited and tends to deteriorate over time. This prompts the use of measured output to adapt the parameters of the model such that it comes closer to reality, a process known as 'history matching' which is currently typically performed on a campaign basis, e.g. once every five years. Furthermore, the matching techniques are usually ad-hoc and not focused on the balanced use of data from various sources. One of the major aims of our research program is to develop techniques to shift from such campaign-based ad-hoc history matching to a near-continuous systematic updating of system models based on data from different sources (e.g. production sensors, time-lapse seismics, passive seismics, remote sensing). To this extent we use and adapt data-assimilation methods that are currently being applied for updating of large-scale numerical models in meteorology and oceanography.
Optimization
A second element of our program is systematic optimization of reservoir production strategies. This involves both optimization in a given configuration, e.g. optimizing the injection and production rates in smart well segments, and in a free configuration, e.g. determining the optimal position of sidetracks or infill wells. The main challenge is to develop techniques that can systematically cope with the large subsurface uncertainties that are typical for the E&P industry. Also here we investigate techniques that have been developed in other industries.
Up- and down scaling
A third element involves scaling (up or down) of system models to the appropriate level of detail. In many cases the controllable subspace of the total system state space is rather small, while at the same time there is a very large unobservable subspace. Similarly the number of model parameters is typically much too large to be identifiable from the available measurements. This offers possibilities for reparameterization, model-order reduction and control-relevant upscaling. Modeling to a level of detail that can neither be observed nor controlled is at best wasted effort, but, worse, may lead to wrong results. We employ various system-theoretical reduction techniques leading to reduced-order models.
Integration and testing
We foresee that the major value of 'smart fields' technology will be in the combined 'closed-loop' use of the optimization, dat assimilation and scaling techniques. The underlying hypothesis is that
βIt will be possible to significantly increase life-cycle value by changing reservoir management from a batch-type to a near-continuous model-based controlled activity.β
Where possible concepts and algorithms are tested on real assets. In addition, we use 'virtual assets', i.e. independent models reflecting reality and including imperfections such as measurement noise, to test the long term (multiple years) implications of the integrated use of data assimilation, optimization and scaling methods.
References
- Model-based optimization of oil and gas production
Tutorial presented at the IPAM Long Workshop on Computational Issues in Oil Field Applications. Institute for Pure and Applied Mathematics (IPAM) at the University of California Los Angeles (UCLA), USA (20 March-9 June 2017). - Control and optimization of sub-surface flow
Keynote speech at the SIAM Conference on Computational Science and Engineering, Boston, Massachusetts, USA, 25 February - 1 March, 2013. - Jansen, J.D., Bosgra, O.H. and van den Hof, P.M.J.
Model-based control of multiphase flow in subsurface oil reservoirs
Journal of Process Control (2008) 18 (9) 846-855. https://doi.org/10.1016/j.jprocont.2008.06.011 - Jansen, J.D., Douma, S.G., Brouwer, D.R., Van den Hof, P.M.J., Bosgra, O.H. and Heemink, A.W.
Closed-loop reservoir management
Paper SPE 119098 presented at the 2009 SPE Reservoir Simulation Symposium, The Woodlands, USA, 2-4- February. https://doi.org/10.2118/119098-MS