This thesis aims to reduce the order of the models of both the four-rotor turbine and the tower, where the four-rotor turbine was developed by the Vestas company. The tower model was reduced from 20 to 17 states using Balanced Truncation and from 20 to 11 states using Matched DC Gain. Then the nonlinear four-rotor turbine model was linearized. The linearized four-rotor model was reduced from 20 to 4 states using the Matched DC Gain method. Two goals should be achieved to reduce the cost of operating the four-rotor turbine in a safe condition. The first goal is to maximize or track the power produced by the turbine. The second goal is to reduce the load on the tower. Thus, there was a need to use modern controllers, such as Model Predictive Control. The Model Predictive Control was developed, and the real-time running were tested. Moreover, the nonlinear Model Predictive Control and linear Model Predictive Control were tested. Different scenarioswere used, starting from the full-order nonlinear four-rotor turbine with full-order tower models. Then reduced-order tower used with nonlinear four-rotor turbine, full-order linearized four-rotor turbine and reduced-order turbine. The power tracking, load reduction and real-time optimization were achieved in most scenarios of linear Model Predictive Control. Whereas nonlinear Model Predictive Control could not run in real-time, and the power tracking and load reduction were below requirements.
Keywords: Model reduction, Model Predictive Control, real-time, power tracking, loadreduction, Balanced Truncation, Matched DC gain.