Sensorless model-based MPPT of micro-wind energy conversion systems using topological Artificial Neural Networks

The power conversion stage and control system of Micro-Wind Energy Conversion Systems (µWECSs) are usually very simple to reduce cost and maintenance. However, such systems exhibit low yielded power and energy, as well as startup issues and abrupt stoppage in response to large wind speed reductions. To solve such problems, this paper proposes to replace their maximum power tracking algorithm with a model-based one, and to implement an auto restart feature. The additional sensors required by the model-based approach (i.e., the anemometer and the encoder) are implemented as virtual sensors to obtain the advantages of a sensorless system such as reduced cost and maintenance, and improved reliability. The virtual anemometer is implemented using a topological Artificial Neural Network, namely a Growing Neural Gas network, whereas the virtual encoder is obtained using a zero-crossing detector. With the proposed approach, no changes are required to the power conversion system of existing µWECSs. The performance of the proposed system is verified experimentally in comparison with a commercial system. The obtained results show improved performance and validate the study.