Global warming is a major consequence of high carbon dioxide emissions due to the burning of fossil fuels. In addition, the use of fossil fuels also emits mercury, sulfur dioxide, nitrogen oxides, and particulate matter into the air and water leading to many health problems. These factors in combination with the depletion of fossil fuel motivate the requirement for low-carbon and renewable energy sources. 

Wind energy takes an important role in the transformation of the global energy system towards clean and sustainable sources. The main development of wind energy technology in recent decades is the growth of Wind Turbine (WT) size motivated by economic factors. The larger turbine size helps to increase power output and energy efficiency, however, it leads to challenges in wind turbine operation and maintenance. Larger and more flexible turbines experience higher mechanical stress on the turbine components such as gearboxes, blades, and towers. These structural loads may lead to early failure limiting the turbine size and performances.

To further reduce the cost of wind energy, advanced control approaches are developed focusing on power maximization, structural load mitigation, lifetime extension, and reliability improvement ultimately reduce the cost of wind energy. This multi-objective problem is difficult to solve due to design conflicts. The optimal trade-off between goals is varying and depends on actual operating situations such as on-site wind characteristics, system aging, and grid requirements.

Advanced control approaches are applied for utility-scale WTs to maximize power production and reduce structural loads. When the structural loads are considered, wind turbines become Multi-Input Multi-Output (MIMO) systems. Because of the coupling between control inputs and outputs, traditional Single-Input Single-Output (SISO) controllers are difficult to design and not suitable for such systems. Multi-input multi-output control approaches consider system internal connections so they can realize multiple objectives simultaneously. Multi-objective advanced MIMO control algorithms reduce the loads while maximizing the power generation. Related control approaches need to be robust and able to reduce the effects of unknown variable wind speed disturbances and modeling errors. 

Load mitigation helps to expand the turbine lifetime, reduce the maintenance cost, and allows to build larger WTs. However, load reduction often comes with the consequence of decreasing power production and increasing blade pitch activities. To define an optimal compromise with these contrary goals, complete knowledge about various elements affecting control performance is required. Besides, the contribution of each aspect to the addressed conflicting objectives as load mitigation, and energy maximization, need to be evaluated by suitable measures. 

Modern utility-scale wind turbines are equipped with numerous sensors providing useful information about turbine components operation status. With the huge development of computation capability and big data analytics techniques, the turbine performance and state-of-health information could be obtained and evaluated through historical logged data using Prognostics and Health Management (PHM) systems. The information aids the optimal operation and maintenance of wind energy systems. In recent years, the integration of state-of-health information into the closed-loop control system begins to attract the attention of the wind energy researcher community. Controllers are adapted based on current and future aging behaviors optimizing the trade-off between service life expansion and power production maximization. 

This thesis develops multi-objectives MIMO control strategies to maximize power production, reduces fatigue loading, and improves the reliability of large-scale WTs. Firstly, the thesis proposes novel measures based on time-series historical data obtained from wind turbines, such as blades/tower bending moments and rotor/generator speed, and the covariance of the data to assess the overall control performance of a wind turbine. New parameters defining the relation between control goals are introduced, which add new measures for controller assessment and design. The measures are able to express multi control objectives graphically and related mathematical values. Secondly, robust control algorithms regulating the generator power, and reducing fatigue loads are developed considering wind disturbances and model errors due to the use of linearized models and unmodeled dynamics. The approaches utilize an unknown input observer scheme to estimate wind disturbances. The WT nonlinearities and unmodeled dynamics are assumed as additive inputs so they also can be estimated by the observer. The effects of unknown inputs including wind disturbance, nonlinearities, and unmodeled dynamics are accommodated using suitable feed-forward controllers. The overall control system including observers and controllers are optimized by minimizing the H-infinity norm of the generalized system with uncertainties. The optimization problem defines optimal control parameters guaranteeing both performance and robustness. Finally, a PHM module providing current and future health information is integrated into the control loop to define the optimal balance of the trade-off between power production and loads mitigation. The PHM module predicts the Remaining Useful Life (RUL) of the system in real-time, so the lifetime of the WTs can be controlled to ensure the turbine survivability to the next maintenance schedule. A novel adaptive lifetime control scheme using RUL prediction is proposed to avoid unwanted failures. The proposed control strategy provides an optimal balance between maximize power production and reduce fatigue loading objectives. The reliability and lifetime of the WTs are controlled guaranteeing the systems reach designed lifetime, reducing unscheduled maintenance cost.