Information technologies and innovative hardware solutions provide novel opportunities for systematic reservoir management. Downhole sensors and remotely operated wells in combination with advanced software for data processing combined with oil reservoir models allow for monitoring and control of reservoir fluid dynamics. Application of advanced optimization and control techniques give improved production [1].

However, reservoir management includes complex decision making on different hierarchical levels. Time scales range from real-time production optimization on an hourly or daily basis to long-term decisions on drainage strategies where the life-time of an asset comes into play. The complexity of decision-making for a large production system exceeds by far the magnitude in which an all-encompassing global optimization problem can be formulated and solved. Although model-based optimization algorithms are utilized to some degree, the state of the art in operations of oil and gas fields still relies on heuristics. Valuable resources are lost because of poor coordination [2].

We describe a novel approach to distributed optimization and control of offshore oil production systems. The model incorporates a complex pipeline network. Oil and gas production systems are represented as a network of connected hierarchical structures of sub sea wells, manifolds and clusters. We consider multiphase flow of water, gas, and oil in the pipelines, and account for discrete switching and typical inflow characteristics of the sub sea wells. Network methods based on variational calculus provide a modeling framework for decentralized optimization and control. Conservation laws and the second law of thermodynamics combined with the passivity theory of nonlinear control lead to conditions for stability and optimality. We describe interconnections in networks through matrix representations that capture a network's topology. Control strategies are derived from the model, and stability and convergence to the optimal solution follows from the passivity conditions. The proposed distributed controller network can be seen as a special case of a Multi Agent System [3].

[1] Jansen, J.D., (2007), Model-based control of subsurface flow, DYCOPS 2007, Cancun, Mexico.

[2] Kosmidis, V. D., Perkins, J. D., Pistikopoulos, E. N. (2005), A mixed integer optimization formulation for the well scheduling problem on petroleum fields, Comput. Chem. Eng., 29 1523

[3] Rawlings, J.B., Stewart, B.T. (2007), Coordinating Multiple Optimization-based Controllers: New Opportunities and Challenges, DYCOPS 2007, Cancun, Mexico.