Once a petroleum asset is in production, strategies for reser-voir management and production optimization are central in all field operations. The overarching goals of these strategies are to maximize the return value of investments in terms of recovery of hydrocarbon reserves. In addition, in later years, minimizing the environmental footprint of the activities is taking a more central place. It is also important to take into account the safety implications of reservoir management and production optimization decisions.

There is huge potential in taking a structured and analytical approach to reservoir management and production optimi-zation. Analytical approaches refer to systematic processes to reach decisions based on a combination of models and data analytics.

Reservoir management and production optimization involves decision-making on several horizons, from the day-to-day decisions regarding choke settings, artificial lift, etc., to longer-term decisions regarding EOR rates and chemical additives, all the way to deciding new production and injection wells. Traditionally, responsibilities are divided between production engineers for the shorter-term deci-sions, and reservoir engineers for the longer-term decisions.

Challenges and opportunities

There are a multitude of challenges in decision-making for reservoir management and production optimization. The majority of these challenges are rooted in the large uncertainties related to reservoir properties and wellbore/pipeline flow. Most of these uncertainties are related to the limited information that is available regarding the subsur-face, but there are also typically large uncertainties and/or noise in the data that are actually available. The uncertainty also has a timescale aspect to it, as indicated in the figure below.

The split in responsibilities between reservoir engineers and production engineers may imply that there is a consid-erable upside in handling the different timescales simulta-neously. There is a need for improved and new tools and methods that help engineers handling these challenges.

Research strategy

The BRU21 approach to reservoir management and production optimization is based on these major digital/automation technology disciplines: Machine learning and hybrid (mixed data-driven and physics-based) modelling, optimization and decision-making systems, and autono-mous systems and automatic control.

Some of the above-mentioned potential can be unlocked by production optimization tasks with short-term hori-zons, e.g. finding the optimum settings of well chokes and artificial lift. Finding these optimum settings requires not only understanding the reservoir potential through reservoir models, but also continuously leveraging on the information available in past and present production data. To be able to achieve this, machine learning from production data must be combined with physical reservoir and production network models through hybrid modelling to generate fit-for-purpose predictive analytics and optimization models.

Decision strategies for processes occurring deep in the reser-voir are based on longer-term horizons and relatively sparse amounts of measured data. Simulation models for fluid flow in full field reservoirs are known for their high computational demand, and workflows for optimization of management strategies tend to be manual due to the complexity of the simulation models. This has left untapped opportunities in auto-mated optimization of reservoir management tasks, e.g. long term well control settings and well placement.