New Hybrid Method Proposal for Wind Speed Prediction: a Case Study of Lüleburgaz

Ümit Kemalettin Terzi, Nevzat Onat, Selcuk Atis


This study proposes a hybrid prediction method where Back Propagation Artificial Neural Network (BP-ANN) and Adaptive-Network Based Fuzzy Inference System (ANFIS) Cascade model are used together to predict the wind speed. At the initial stage, to increase the accuracy of prediction, an additional ANFIS layer is used which is driven with outputs acquired from BP-ANN. Thus, a Sugeno type ANFIS model is used for future prediction and prediction capability of a conventional BP-ANN algorithm is increased. At the second stage, BP-ANN outputs and cascade model outputs are compared to the measured value and a selection criteria algorithm is developed that accepts the output very close to the real value. The proposed model generates a prediction output based on a hybrid operation of the two systems unlike the other study in literature. The proposed model is tested with the data taken from the sample station and the results are compared to real calculation values with regard to various statistical error parameters. Results of the study have shown that the hybrid model generates the closest prediction results to the real values compared to the other models. This flexible algorithm developed to predict the wind speed can be also used in other fields in future research.


Back Propagation ANN, ANFIS, wind speed prediction

Full Text: PDF

Print ISSN: 1392-1649
Online ISSN: 2029-2139