Tentative Technical Agenda

Tuesday, 24 June 2008

Session One — Assimilation of Production and Seismic Data into Reservoir Models

Session Managers: Wen Chen, Chevron, and Akhil Datta-Gupta, Texas A&M University

This session will focus on the integration of production and seismic data into reservoir models. The emphasis will be on the role of history matching on improved performance forecasting and production optimisation as opposed to obtaining a single history-matched model. Field examples will be included in most of the presentations. Some of the topics addressed will be

Data assimilation in other disciplines
Integration of time-lapse seismic for improved reservoir management
Sequential (continuously updating models when new measurements are available) vs. nonsequential/traditional (updating the model and matching all available measurement data using an optimisation algorithm) approaches to history matching
Uncertainty assessment in performance forecasting

Session Two — Life-Cycle Production Optimisation Under Uncertainty, Including Well Positioning

Session Managers: Khalid Aziz, Stanford University, and Greg Walker, BP

The second session will examine the effects of the depletion plan on a field, with the intent of covering both how robust the depletion plan is to the geological uncertainty and how the depletion can generate surveillance information to reduce uncertainty. We will also aim to explore a range of timescales, crossing from real-time optimisation, which may act as a sequence of interference tests, through to longer strategic decisions involving well placement and potential changes in recovery mechanism as the field matures.

How should we optimise future production on the portfolio of models?
How should we simultaneously optimise surveillance acquisition and production to close the loop?
How is the optimisation altered by the frequency of model updating?
How should long-term reservoir management be linked to short-term production optimisation?

The session will emphasise field examples with a combination of uncertainty mitigation strategies through different depletion plans and identification of feedback loops for different types of surveillance.

Wednesday, 25 June 2008

Session Three — Decision Making and the Value of Information

Session Managers: Chris Farmer, Schlumberger/University of Oxford, and Dean Oliver, University of Oklahoma

This session will focus on the various sources of uncertainty, their effect on the decision-making process, and the computation of optimal production controls. Although the optimal control strategy is clearly a function of the unknown reservoir properties, uncertainty in future oil and gas prices will also impact particular economic objectives. Speakers will address the value of information on decision making and methods of reducing uncertainty in reservoir forecasts through improved reservoir monitoring. Specific questions will include when, where, and what data should be collected to optimise decision making and production objectives.

Session Four — System-Theoretical Aspects

Session Managers: Sigurd Ivar Aanonsen, Centre for Integrated Petroleum Research, and Alberto Cominelli, Eni

In closed-loop reservoir management, the number of model parameters to be estimated and the number of controls may both be large. However, basic control theory indicates that reservoir system models often have many unobservable and/or uncontrollable states, i.e., many more states than can be uniquely determined from the outputs, or can be uniquely influenced by the inputs. Similarly, the model parameters are usually not uniquely identifiable from input-output measurements. Model reduction and reparameterisation can be applied to adjust the model complexity to a level that makes the states controllable and observable and the parameters identifiable. This session focuses on theoretical and practical aspects of modern control theory that can impact applications of closed-loop reservoir management, with examples from other industries where control theory applications are routine and from integrated operations in the oil industry.

Presentations will cover the following areas:

Methods for model reduction, regularisation, and reparameterisation
Identification and model-based control of dynamical systems, with applications in industrial processes and reservoir engineering

Thursday, 26 June 2008

Session Five — Presentation of the Benchmark Problem

Session Managers: Rob Arts, TNO, and Geir Naevdal, IRIS

The workshop committee will develop a test case for data assimilation and production optimisation methods on a waterflooded oil reservoir. To this end, they will release a dataset, with the intent that workshop delegates can discuss their history match and production strategies from a common basis, thus providing a reference for future developments in this field. Although the production optimisation step will be done much less often than in true closed-loop reservoir management, the test problem will require application of the same principles necessary for practical implementation of the method. The session will effectively start prior to the ATW. In February 2008, a 3D synthetic dataset will be made available to interested parties. The dataset consists of 100 upscaled realisations of a 3D geological model (in Eclipse format); well-log data from wells with fixed positions; the first 10 years of the production history of the field (including measurement errors); inverted time-lapse seismic data in terms of (uncertain) pressures and saturations; and economic parameters for oil, water, and discount rate.

We would like participants to come up with a history match (either a single matched “best” model or a matched ensemble) based on the available data and an optimal production strategy (without infill drilling) for the next period (20 years). If that strategy is ready to test in April–May 2008, participants are invited to send it to TNO for testing on the “real field” to obtain additional production data over a 10-year period. Using these production data, the participants can update their reservoir model and revise their optimal production strategy for the final period of production. Participants are invited to, but not required to, present their results at the workshop.

To participate and obtain the dataset with instructions, please contact Rob Arts at TNO. (rob.arts@tno.nl).