Extreme learning machine based control of wind turbine with permanent magnet synchronous generator

This dissertation focuses on an extreme learning machine based controller that can perform full-scale control of a wind turbine with permanent magnet synchronous generator. In order to design an intelligent controller, a system modeling should be performed in which the stable operating parameters are selected appropriately. For this purpose to model of a wind turbine, firstly the main-system must be modeled mathematically after that the mathematical information about the wind turbine and the generator are given in Chapter 2. In Chapter 3, the equipment which contains the basis of the system such as back-to-back inverter, maximum power point tracking, phase looked loop, LCL filter and transformer have been examined and modeled. In this section, optimal PI controller design, pulse width modulation techniques, artificial neural network and extreme learning machine are also investigated. In Section 4, the steps for the design of the control system in the wind turbine have been showed. In this section, the determination of optimal maximum power point equation, design of LCL filter and determination of the transfer function are investigated. Then, PI controller design based on the optimal modulus and optimal symmetry criteria were realized. After the basic model was obtained, a new controller are designed by using a single hidden-layer artificial neural network called extreme-learning machine. In order to examine the performance of the neural network trained as controller, both performance indices and speed performance tests are shown. Wind turbine speed, torque, maximum power point tracking curve, d-q currents etc. are presented for the designed wind turbine system in Chapter 5. In addition, on the grid side DC bus voltage ,d-q currents, LCL filter voltages and currents are presented. The results of the extreme learning machine based controller are shown. The obtained results are discussed.