Decoupling control for a magnetic levitation wind turbine using neural network inverse scheme plus model reference adaptive controllers

As one of the well-known renewable energy resources, wind energy, which attracts a larger part of today's total investments, is now playing an increasingly important role around the world.Because the high cost and vulnerability of the gearboxes in wind turbines, the direct-drive wind turbines without gearboxes were replaced with the traditional wind turbines by the manufactures in recent years. But the threshold wind velocity of direct-drive wind turbines is higher than the traditional wind turbines, because of the size and weight of the direct-drive wind turbines. In addition, the working condition of wind turbines is very bad, such as heat, high humidity and impact loading. So the magnetic bearings with well performance of sealing, lubrication, shock resistance and long-lifetime can used in the direct-drive wind turbines to solve these problems mentioned above. The magnetic suspension system is a time variation system with multivariable, close coupling and nonlinear, which consists of some magnetic suspension devices such as magnetic bearings, suspension windings and so on. So the magnetic levitation control technology is difficult and complex in magnetic suspension system. The individual controllers with traditional linear control scheme, i.e., the proportional-integral-differential (PID), were used in magnetic suspension system, ignored the coupling between the variables in system. In order to realize high accuracy and high performance of the magnetic suspension system, the high-performance decoupling control is necessary in suspension system. Many control methods were put forward by some scholars around the world. The control of H∞ extension and Fuzzy were used in vehicle active suspension system. LQG control was used in Low bandwidth active suspension system. The internal model control system is used in bearingless permanent magnet synchronous motor by Sun xiaodong. The neural network inverse (NNI) system technology is advancing fast in different fields. For complex nonlinear systems, the NNI control scheme can play a role in feedback linearization and multi-variable decoupling, and this idea is simple and easy to understand. On the other hand, some uncertainties are inherent in nonlinear systems due to modelling error, loading changing and environment. The model reference adaptive (MRA) control is used in the nonlinear systems, where the tracking signal comes from a set of reference dynamics. So the NNI and MRA control can be adopted in the 5-DOF magnetic suspension system.