Comparative Study of Wind Turbine Identification by Using Neural Networks

In this study, a wind energy conversion system is aimed to be identified by using different identification techniques. The wind speed input data is constructed using the appropriate amplitude and time delay in order to be consistent with real data in the large range of wind turbine operation. The three input parameters including wind speed, pitch angle, and rotor speed and the output power as an output parameter are used for identifying the DFIG wind turbine. Several identifier structures such as one layer MLP, two layers MLP, RBF, NRBF, LOLIMOT, and NARX are employed for the identification purpose. The simulation results demonstrate that the error of train and test data for identified system is so trivial. The RBF and NRBF have the least mean square errors for the test and train data, so that these structures can be considered as the best identification approaches among all for the DFIG wind turbine system identification.