Experimental validation of distributed feedback-based real-time optimization in a gas-lifted oil well rig

This paper considers the problem of steady-state real-time optimization (RTO) of large-scale processes with a common constraint for several units, for example, a shared resource. Such problems are often studied under the context of distributed optimization, where each subsystem is locally optimized for a given shadow price of the shared resource. A central coordinator is then used to coordinate the allocation of the shared resource. In traditional RTO, such a framework would require repeatedly solving the subproblems and the central problems until convergence, which can be computationally expensive. To address this issue, we use a feedback-based distributed RTO scheme based on Lagrangian decomposition, where the local subproblems and the central problems are converted into feedback control problems. That is, by appropriately choosing the controlled variables in each subproblem, the overall process can be asymptotically driven to its optimal operation using feedback controllers. In this paper, we validate this approach using a lab-scale experimental rig that emulates a subsea oil production network, where the common resource is the gas lift that must be optimally allocated among the wells. We also benchmark its performance with a numerical optimization-based RTO scheme.