A High-Performance FPGA-Based Virtual Anemometer for MPPT of Wind Energy Conversion Systems

This paper proposes a fast and compact implementation of a virtual anemometer on a low-cost Field Programmable Gate Array (FPGA) platform. Such an anemometer is to be used within Wind Energy Conversion Systems (WECS) to perform Maximum Power Point Tracking in a non-iterative way, thus reducing dead time and increasing yield. The proposed virtual anemometer relies on a Growing Neural Gas (GNG) Artificial Neural Network with 512 neurons. A major effort is placed on hardware optimization, aiming to achieve the best compromise between computational speed and resource occupation. Furthermore, the slave SPI interface allows a fast communication with the main microcontroller on which the WECS control system is implemented. The resulting design is a high-performance virtual anemometer that can be embedded in WECS control systems with up to 100 kHz bandwidth. The device is designed, synthesized and implemented on a commercial FPGA. Several details of the implementation are discussed, and an experimental validation is performed using input profiles that have been acquired on the field for two different wind turbines.