Özet:

The integration of wind energy into the electricity network is an important topic for economic distribution and the short-term wind speed forecast for the control of the safe operation of wind turbines. The immediate change in wind speed makes the problem more difficult. In this study, a new hybrid model is recommended for the short-term wind speed forecast, with linear autoregressive, AR and nonlinear artificial nerve networks, YSA models being used simultaneously. AR models are statistical methods commonly used in predictive problems. The YSA approach is a approach that works similarly to the nerve networks in the human brain used in the main modeling, prediction and classification problems. In this study, the combination of two powerful methods used in predictive problems has been introduced as a new hybrid approach to determining short-term wind speed. This approach has achieved more successful results in the wind speed forecast compared to the methods only used by the AR method or the method only used by the YSA method. For the purpose of demonstrating the accuracy of the study, the eight-year real hour average wind speed values of the Eskişehir area received from the meteorological station were used. Seven-year wind speed values were used as training data, the remaining one-year values were used for testing purposes. The recommended hybrid approach for different cases has shown that root average square error values RMSE and average absolute error MAE values have achieved lower results than the direct use of AR and YSA methods.

Anahtar Kelimeler: