Real-time optimization with persistent parameter adaptation applied to experimental rig

Real-time optimization (RTO) is a steady-state model-based method used for optimizing process operation in chemical plants. The most common implementation, two-step RTO (TS-RTO), updates the steady-state model parameters in the first step, and optimizes this model in the second step. It has a major drawback, which is the need to wait for steady-state. If data from transient periods is directly used for updating the steady-state model parameters, the production optimization results will most likely be sub-optimal, decreasing the benefits. This becomes even more acute if the system is constantly affected by disturbances and has long settling times. Matias and Le Roux [2018] proposed a TS-RTO variant that uses a dynamic estimator to update the steady-state model parameters, which was named real-time optimization with persistent parameter adaptation (ROPA). By using dynamic estimation, it ensures that the model is always updated to the plant and the steady-state optimization can be scheduled at any desired rate without needing to wait for steady-state. This hybrid approach has been successfully tested in simulations. In this paper, we show its first implementation in a lab-scale rig, which emulates a subsea oil well network. The results show that the hybrid approach enables an increase in the optimization frequency and a decrease in the optimization results variability, improving the overall economic performance when compared to the TS-RTO implementation.