This program shall support the following Program 2 objectives

1. Embed model-based decision-support tools for RTPO into new work processes
2. Develop methods and prototype technology for RTPO – including the value chain optimization problem
• Methods to support real-time production optimization. The goal is to improve efficiency and predictability of the optimization methodology itself, and improve flexibility in terms of the types of proxy models that can be used.
• Optimization technology that allows for decomposition of large problems into parts. The goal is to support a divide-and-conquer strategy which lends itself to common workflow patterns in engineering.
• Develop proxy models from high-fidelity models. The goal is to create fit-for-purpose models along the production chain as a basis for integrated optimization.
3. Communicate results with industrial partners
4. Publish and present results.
5. Educate PhDs and master students.
6. Establish strong collaboration with Carnegie Mellon University

Computational integration

Computational IO (system integration) with model-based optimization is an exciting, new, and challenging approach to integrated operations. The approach is not currently used by the industry on a full scale, for many obvious reasons, but the industry has given a clear indication that this is an ultimate goal within IO. The concept is to (electronically) pipe together operations from far upstream to far downstream, including reservoir, production, flow assurance, processing, market, and full-scale economics. Local (operation) and global optimization is a prerequisite.

Integrated operations can rightfully be challenged by the divide-and-conquer strategy; a sound engineering design principle which has survived ever since complex system came into making. This program will explore approaches to obtain the best from both worlds, for instance by looking into methods in which local solutions give rise to close-to-optimal operation.

Proxy-models

Proxy models are necessary to solve the value chain optimization problem. In practice this means developing simple models directly or extracting simplified models from detailed simulators like Eclipse and HYSIS. Both physics-based as well as data-driven models will be investigated. The simplified models are fit-for-purpose models, i.e.

• they are modelled with a particular goal in mind (RTPO),
• their complexity gives a tractable optimization problem, and
• the models are sustainable over time

The one-fits-all paradigm will to some degree be sacrificed to promote simplicity. An example of this is production optimization during the ramp-up phase of an asset as compared to the decline phase may be two very different problems requiring different models.

The industry is reluctant towards the use of dynamic models. Time dependency introduces complexity, but also opportunities. A long pipeline may e.g. be used actively as a buffer during some transient or some contingency. Dynamic models will hence be included in the research.

Optimization – decentralized approach

Technologies for tactical planning need to take into account all relevant constraints and interactions in the production systems. Problems and decisions will vary during the life cycles of an asset, e.g. in a green-field as opposed to a brown-field. There are interesting opportunities in optimization by taking advantage of recent developments in convex programming, mixed-integer programming and methods for decomposing large scale problems. It is appropriate to design research tasks developing and analyzing these methods to suit the challenges of real-time production optimization. Important issues are how to model nonlinearities, and how to include uncertainties in a meaningful manner.The first test bed for research results is the Troll Oil field.

Work processes

Through visits in participating oil companies, document studies, interviews and investigation of work- and decision making processes, we seek to identify key workflow-challenges related to production optimization. These studies will provide input and context to the development of technology and methods within the program. We will attempt to convert successful results into decision support tools that can be embedded into operational use.

Pilot case

Several pilot studies will be used for RTPO. The first is the Troll west oil rim. The RTPO problem is highly complex and it suits well as a test bed for optimizations methods, integrated optimization and alternative proxy models. Further, the organization is positive towards cooperation.Other case studies will be launched.

Cooperation with IO-center industrial partners

The Troll west oil rim is operated by StatoilHydro is the first field case for Program 2.1. This case will be important in Program 2.2 in 2008-2009. Cooperation has been established with FMC on proxy-models and optimization. Test cases will be defined. This activity will be developed further in 2008-2011.Cooperation with PERA/Petrostreamz has been established. Focus is on proxy-models and optimization, and test cases will be defined. This activity will be developed further in 2008-2011.Results will be disseminated through workshops and publications.

Cooperation with IO-center academic partners

Important research on real-time decision making is performed at the Center for Advanced Process Decision making (CAPD) at Carnegie Mellon University (CMU). The focus there has been on the downstream industry, but methods and technology are clearly highly relevant. This includes model reduction, network models and methods for optimization. The research in Program 2.2 will seek close cooperation with this research centre. A concrete step is a pair pair of PhD students at NTNU and CMU working on network optimization and decision making in distributed organizations. Resources will be allocated for researchers at the cooperating centre.

Cooperation with other IO-center programs

This program will cooperate with program 1 and 4.

• This program will benefit from cooperation with program 3.3 Integrated Planning. Program 3.3 addresses complementary planning activities to RTPO. Hence, exploring a common framework in terms of methods and work processes makes sense. No immediate activity will be prioritized.
• This program will benefit from cooperation with program 4.2a Decision-making in multidisciplinary and distributed teams and 4.2c Information design for improved decision-making. The reason is the fact that multidisciplinary teams are critical in RTPO, especially when a large part of the value chain is covered. A cooperation has been initiated within the framework of the Shell Draugen study.

There is a need to explore the link between optimization and decision-making on different time-scales as stated in objective 3 Enhance the understanding of the link between production and field optimization. Program 2.1 and 2.2 will be designed to explore the link between long term and short term decision. One approach is a direct approach where the long term and short term objectives are included into one large problem. An alternative, and probably superior approach, is to divide the problem into parts. Then, the long term strategy feeds the short term tactics (i.e. the short term optimization problem) while the short term results is fed into the long term strategy. One such example is the relationship between (long term) water management and (short term) breakthrough of gas.