Universal Maximum Power Extraction Controller for Wind Energy Conversion Systems Using Deep Belief Neural Network

Over the previous years, there has been a surge in need for wind turbines. Consequently, this paper aims to propose a Deep Belief Neural Network (DBNN) that preserves the Maximum Power Extraction's (MPE's) benefits while allowing the output power to be limited with substantially less complexity in the control loop. The suggested technique uses a deep belief neural network to learn the wind turbine's nonlinear aerodynamics. It is generalized to cover a wide range of wind turbine sizes and operating conditions to precisely track their maximum power trajectories. Also, the proposed method rewrites the machine model in order to run Wind Energy Conversion Systems (WECSs) under the MPE technique; which takes into account the radius and pitch angle of wind turbine blades, wind speed, air temperature, and power demand for any permanent magnet synchronous generators types (IPMSG or SPMSG). Moreover, the proposed methodology is considered a plug-and-play technique while it does not need any tuning measures. The suggested DBNN MPE controller's power tracking achievement is examined in various operating conditions using a set of experimental and simulation tests for different generator types and sizes. Finally, the findings are compared to the well-known verified technique to validate its generalized productiveness.