Anti-Slug Control with Non-Linear State Estimation

In offshore production, the two-phase mixture of oil and gas is transported from the seabed oil wells to the surface facilities by pipelines and risers. The two-phase flow can have different flow regimes, where severe slugging is one undesirable flow regime and an effective solution is needed to prevent it. The recommended solution is active control of the top-side choke valve. Previously, controllability analysis is done of two-phase flow in a 4-state pipeline-riser model. This analyze concludes that the best way to control the choke valve is by using the subsea pressure measurement combined with topside flow measurement. However, the subsea pressure might be difficult to measure correctly because the pipeline is placed under tough conditions, hundredth or even thousands of meters under sea level. One possibility is to combine topside pressure with topside flow measurement and use for estimation of states or other sub-sea measurements that are normally not available. Simulation studies are done in MATLAB of different anti-slug control solutions. Linear Kalman filter, extended Kalman filter (EKF) and unscented Kalman filter (UKF) are used for state estimation and combined with controllers such as PI, LQR and MPC. The input to the system is flow rate of gas and liquid, and the nominal choke opening. The input disturbance to the process is change in the flow rate of gas and liquid imitating slug flow.As expected, when only topside measurements are used, because of the highly nonlinear system dynamics the linear Kalman filter fails in stabilizing the system. The EKF works good when the system has low input disturbance, while the UKF is the best nonlinear filter when the system has high input disturbance. However, when the nominal choke opening is increased, the UKF combined with a controller fails. The LQR controller combined with UKF shows slightly better results than the PI and MPC controller combined with the same filter for state estimation. There is also potential in using the high-gain observer in control strategies