Neural Network Based Wind Speed Sensorless MPPT Controller for Variable Speed Wind Energy Conversion Systems J. S. Thongam and P. Bouchard R. Beguenane I. Fofana Department of Renewable Energy Systems Department of ECE Department of Applied Sciences STAS Inc. Royal Military College of Canada University of Quebec at Chicoutimi Chicoutimi, QC, Canada PO BOX Kingston, ON, Canada Chicoutimi, QC, Canada {thongam.js & bouchard.pierre}@stas.com Rachid.Beguenane@rmc.ca ifofana@uqac.ca Abstract - A wind speed sensorless neural network (NN) based maximum power point tracking (MPPT) control algorithm for variable speed wind energy conversion system (WECS) is proposed. The proposed method is developed using Jordan type recurrent NN which is trained online using back-propagation. The algorithm, without requiring the knowledge of wind speed, air density or turbine parameters, generates at its output the optimum speed command for the speed control loop of the vector controlled machine side converter control system using only the instantaneous power as its input. The output of the NN is fed into a state input after a unit step delay completing the Jordan type recurrent neural network. The proposed concept is analyzed in a grid connected direct drive variable speed permanent magnet synchronous generator (PMSG) WECS with a back-to-back frequency converter. Vector control of the grid side converter is realized in the grid voltage vector reference frame. Simulation is carried out in order to verify the performance of the proposed controller. Index Terms— Permanent magnet synchronous generator, wind energy conversion system, vector control, maximum power point tracking control, wind speed sensorless. I. INTRODUCTION Wind energy conversion systems have become the focal point of research efforts made in the area of renewable energy systems. This has become possible due to the rapid advances in the size of wind generators as well as the developments in power electronics and their application in optimum wind energy extraction. In recent years, fixed speed wind energy conversion systems, due to poor energy capture, stress in mechanical parts and poor power quality have given way to variable speed systems. These systems can be controlled in order to enable the turbine to operate at its maximum power coefficient over a wide range of wind speeds, obtaining a larger energy capture from the wind in addition to having reduced mechanical stress and aerodynamic noise [1-4]. One of the problems associated with some of the variable speed systems of today is the presence of gearbox, coupling the wind turbine to the generator. This mechanical element suffers from considerable faults and increases maintenance expenses and hence direct drive WECS are becoming increasingly popular. Wind power, even though is an abundant source of energy, the power that can be obtained from it changes throughout the day as wind speed changes continually. The maximum power which a wind turbine can deliver at a certain wind speed depends upon certain optimum value of speed at which the rotor rotates. Extracting maximum possible power from the available wind power is of utmost importance, because, only when we do it can have efficient use of equipment and available energy source; therefore, MPPT control is an active research area. In order to have maximum possible power the turbine should always operate at optimum tip speed ratio. This is possible by operating the turbine at the optimal speed of rotation where the tip speed ratio is optimum. Wind speed sensor normally used in most of the WECS [5- 7] for implementing MPPT control algorithm reduces the reliability of the WECS in addition to inaccuracies in measuring wind speed. Therefore, some MPPT control methods estimate the wind speed [8-12]; however, these methods require the knowledge of air density and mechanical parameters of the WECS. Such methods, requiring turbine generator characteristics result in custom-design software tailored for individual wind turbines. Air density, on the other hand, depends upon climatic conditions and may vary considerably over various seasons. Therefore, a lot of research efforts are made to develop wind speed sensorless MPPT controller which does not require the knowledge of air density and turbine mechanical parameters [13-18]. In [13] MPPT control is achieved using stator frequency derivative and power mapping technique. The maximum power curves for power mapping are established by running several simulations or offline experiments at various wind speeds. Hill climb search (HCS) control method is used in [14-19] in order to track the peak power points. In [14] fuzzy logic MPPT controller is presented wherein the controller uses active power as the input and reference speed signal is generated at its output. Optimum power search algorithm is proposed in [15] which uses the fact that dP o /dω=0 at peak power point. The algorithm dynamically modifies the speed command in accordance with the magnitude and direction of change of active power. Hill climb search algorithm, proposed in [16] for tracking the peak power point, uses the principle of search-remember-reuse process. The method uses memory for storing maximum power points obtained during training process which are used later for tracking peak power points. Sliding mode extremum seeking control is proposed in [17] for maximum power point tracking. In [18], a simple MPPT controller is presented wherein a smooth tracking is