Lately, wind energy conversion systems are one of the main renewable energy sources in the world. By 2017 total installed power in wind energy systems was 487 GW, with estimate that this number will reach 760 GW by 2020. Due to this fact, there are numerous scientific research projects with focus on improving wind energy conversion systems. Wind energy conversion system consists of a wind turbine, gearbox (can be omitted in some configurations), electric generator, (partial or full scale) back-to-back converter and a filter between the converter and the electrical grid. Due to low cost and simplicity, the leading solution includes doubly-fed induction generator with a partial scale (around 30% of generator nominal power) back-to-back converter. However, many scientific papers indicate that solution with a permanent magnet generator and full scale back-to-back converter will take over market in the near future. This configuration allows full range speed control of the wind turbine which results in maximum power extraction from the wind. The main advantages of permanent magnet generators are high efficiency coefficient, high power density, high power factor (magnetic flux is produced by permanent magnets on the rotor) and wide speed range. Depending on the construction, there are surface and interior permanent magnet generators. In a case of surface permanent magnet generator there is magnetic symmetry in direct and quadrature axis which results in fewer generator parameters and simpler control algorithms. On the other hand, if permanent magnets are placed inside rotor, there is magnetic asymmetry in direct and quadrature axis which results in additional, reluctance torque. This type of generator also has wider speed range compared to surface permanent magnet generators but more complex control algorithms are required. Scientific focus of this PhD thesis are advanced control algorithms of interior permanent magnet generators for application in wind energy conversion systems. Narrow focus of research is generator, generator-side converter and its digital signal processor in which developed control algorithms were implemented. Advanced control algorithms include control structures which provide various advantages and improvements over standard control structures such as rotor field oriented control or direct torque control. The main guidelines for research are sensorless control structures for high power generators including flying start, model predictive control of two-level converter and flux weakening control structures. In those specified areas new control structures and algorithms were developed which represent scientific contribution of this PhD thesis. According to some researchers 14% of wind system failures are related to sensor faults. The main motive for developing sensorless control techniques is elimination of sensors and improvement of robustness of the overall wind system. However, most scientific papers are focused on developing rotor speed and angle estimators, with assumption that machine is started from standstill which is not a case of a wind generator. In a case of a wind generator, the generator-side converter is started when wind speed reaches cut-in speed. In such a case, prior to starting a generator-side converter, the control structure should have already determined rotor speed, rotor angle and induced voltage of the generator. Otherwise, significant inrush current occurs which causes mechanical stroke which could be fatal for the whole wind system. Only a few scientific papers and patents analyse switching on a converter during rotation of a permanent magnet machine. Also, control structures developed in those papers are often verified on small power machines with a high value of stator resistance which prevents significant inrush current. In this PhD thesis focus is on developing control structure which incorporates flying start in sensorless control structure for the whole operating range of the high power wind generator. With development of digital signal processors model predictive control has become a promising alternative to standard control structures based on PI or hysteresis controllers. This control principle is based on prediction of future system states using mathematical model of the system. Using predicted future states and taking system and control input constraints into account, the minimum of the defined cost function is found and optimal control input is applied to the system. Model predictive control algorithms applied to power converters are distinguished whether pulse-width modulation is used (continuous control set) or not (finite control set). Finite control set model predictive control algorithms are analysed in this PhD thesis. The main advantage of this algorithm group is better dynamic performance compared to algorithms which are based on pulse-width modulation. In wind systems high dynamic performance is not crucial due to high inertia of wind turbine. However, finite control set algorithms also enable minimizing switching losses of the power converter with proper definition of the cost function. This aspect, along with ensuring stability and recursive feasibility of the algorithm, is the main focus of research in the PhD thesis. One of the requirements in wind generator control system is ensuring operation on the voltage limit, i.e. flux weakening. Rotor speed increases due to increase in wind speed which results in increased induced voltage at generator terminals. If generator voltage exceeds maximum permissible value defined by DC link voltage and used pulse-width modulation technique, the control system becomes unstable. Additional control structure, usually based on generation of negative direct axis current which reduces generator voltage below maximum permissible value, is required. In most scientific papers control systems with PI speed controller and operating point on intersection of current and voltage limit are analysed. However, in case of wind system analysed in this PhD thesis, the controlled variable is torque, so modifications to existing control structures are required. Additionally, in a case of interior permanent magnet generator, standard flux weakening structures cause increase of reluctance torque which results in suboptimal operating point. The research focus in the PhD thesis is robust flux weakening structure which is compatible with standard torque control of wind generator but also enables operation on voltage limit with successful torque reference tracking. The PhD thesis contains following chapters: Chapter 1. - Introduction. In this chapter scientific problem is defined and motivation behind it is given. Also, scientific contributions of the PhD thesis are emphasized and structure of the thesis is outlined. Chapter 2. - Wind energy conversion systems. In this chapter operating principle of a wind energy conversion system is described. Depending on the type of electric generator and power electronic components, four main types of wind energy conversion systems are distinguished. The emphasis is given to the fourth type, specifically a configuration with permanent magnet generator and a back-to-back converter. The mathematical model of permanent magnet machine in different coordinate systems is derived. Standard control structures for generator-side converter (rotor field oriented control with MTPA algorithm) and grid-side converter (voltage oriented control) are explained. Chapter 3. - Laboratory setup. In this chapter experimental setup used for verification of developed algorithms is described. The setup consists of a permanent magnet generator, back-to-back converter and an induction motor which emulates wind turbine operation. Rated data for all aforementioned components along with description of measuring equipment (e.g. torque and current measurement) is given. Digital signal processor of generator- and grid-side converters is described as well. Also, main principles of graphical programming tool (AlgoCad) and supervisory tool (r_parnad) are explained. Chapter 4. - Sensorless control. This chapter is related to development of control structure without rotor position measuring device (position sensor). The proposed control structure consists of flying start and synchronisation phase. During flying start induced voltage at generator terminals is measured and phase locked loop is employed to estimate rotor speed, rotor angle and induced voltage. Once the estimation is complete, generator-side converter is switched on and synchronisation phase is started. During continuous operation estimation of rotor speed and angle is carried out by estimation of magnetic flux and phase locked loop. Simulation and experimental results are given for switching on generator-side converter without flying start process (comparison to standard algorithms applied to motors), flying start and synchronisation phase of the developed sensorless structure along with change of torque reference and rotor speed. Chapter 5. - Model predictive control of currents. In this chapter different model predictive control algorithms applied to electric machines and power converters are analysed, with emphasis on finite control set model predictive control. Basic principles of model predictive control are given. Mathematical model of permanent magnet generator and generator-side converter with and without considering dead time was derived. The developed model predictive direct current control algorithm consists of two modes, namely transient and steady state. By choosing parameters for penalizing current control error and switching state change, trade-off between current ripple and reduced switching losses can be achieved. The stability and recursive feasibility are guaranteed by adding flexible control Lyapunov function constraint. Adding flexibility to standard control Lyapunov function provides less conservative optimal control action which enables penalizing switching losses even during transient state. Simulation results are given for different sizes of control invariant set, comparison of standard and flexible control Lyapunov function, analysis of different values in penalisation matrices, analysis of converter dead time impact and analysis of generator parameters variation impact. Experimental results were given for switching on the generator-side converter with developed algorithm at different rotor speed values. Analysis of different values in penalisation matrices and analysis of compensation of digital system computational delay were given. Chapter 6. - Torque control in flux weakening region. This chapter is related to development of flux weakening structure for permanent magnet generators. Current and voltage constraints of a permanent magnet generator are explained. The developed flux weakening structure is based on coordinated operation of voltage and torque PI controllers. While the voltage PI controller keeps voltage on maximum permissible value, the torque PI controller prevents increase of reluctance torque and ensures correct tracking of torque reference. Trajectory of current in flux weakening region is also described. Experimental results are given for entering and exiting flux weakening operation with and without torque PI controller at different torque reference and rotor speed values. Chapter 7. - Conclusion. This chapter contains conclusion of the research and emphasizes scientific contributions of the PhD thesis. Simulation and experimental results show that the developed control structures for sensorless control with flying start, model predictive direct current control and flux weakening achieve improvement over existing algorithms which can be found in scientific literature. Flying start concept ensures switching on power converter with negligible inrush current while sensorless control structure enables operation within entire operating area of the generator without rotor position and speed sensor. Model predictive direct current control algorithm provides possibility to achieve trade-off between acceptable current ripple and reduction in generator-side converter switching losses. Flux weakening structure ensures operation on voltage limit while simultaneously tracking torque reference value. Low computation burden of the developed algorithms indicates possibility of their utilisation in industry applications.