A Novel Model Predictive Control Formulation for Wave Energy Converters based on the Reactive Rollout Method

In the wave energy converter (WEC) optimal control problem, the well-developed model predictive control (MPC) faces the trade-off between energy extraction performance and online computation burden when choosing the length of optimization horizon. In this work, a novel MPC problem formulation is proposed, which includes a terminal value function and a terminal constraint set. The terminal value function is derived by further calculating the system trajectory beyond the optimization horizon based on a pre-chosen "rollout" policy, which incorporates more future information and improves energy extraction. The linear reactive control is chosen as the rollout policy, under which a policy improvement theorem of the WEC system is derived. The terminal constraint set is a simple control invariant set that ensures the recursive feasibility of MPC. Extensive simulations show that the proposed reactive rollout MPC can achieve close-to-optimal performance with a very short optimization horizon, reducing the online computation burden significantly.