Discrete Model Identification of a wind turbine: a case study of Nordtank NTK 330F

Resumo

The renewable energies are in constant evolution for the sake of the necessity of finding alternatives to
fossil fuels derived energy. One of these alternatives are the wind farms, offshore and/or onshore, that provides
electricity through the wind force. The horizontal axis turbines are commonly seen on this farms, usually with 3
blades, because of its cost-benefit. However, even though there are equations well defined regarding the behavior
of this machines, some environmental variables (such as terrain, height and wind wakes) can add a nonlinear and
random factor to the energy conversion. Thereby, the use of system identification shows itself useful to forecast
the behavior of this system. In this work, a database from Nørrekær Wind Farm, located on Denmark, is exploited
by identification techniques, in order to estimate nonlinear auto-regressive with exogenous inputs (NARX) and
auto-regressive with exogenous Inputs (ARX) models. To calculate them properly, three methods will be used

and compared: Classical Gram-Schmidt (CGS), Modified Gram-Schmidt (MGS) and Householder-based QR-
Decomposition with Column Pivoting. All the methods provided results closer to real data, although MGS and HT

models were slightly more accurate.