عنوان مقاله [English]
In this paper at the first step, the present weather of Metar reports in the Mehrabad synoptic station has been studied and identified the period in which the most instability has occurred and formed the gusty wind. This period is January to June every year. And we used data of 2013 in this study. Then, the all data of selected period, except the data of Gusty wind direction and speed have been normalized to interval 0.1 – 0.9. The 60% of the data has been considered for training, 20% for the test and also 20% for the validation. The related features to Gusty wind direction and speed were selected from between 58 features recorded by 3 sensors located on the runway. The Mehrabad runway has 4000 meters long and 45 meters wide and eastward to the west. The sensor number 29 locates at the eastern end of band, a sensor with a number 11 at the western edge of the band, and the mid sensor located in the middle of band which its distance from band is 600 meters to the north direction.
The feature selection methods in this study are mutual information (MI) with the Maximum-Relevance Minimum-Redundancy criterion (filter type) and Sequential Floating Forward Selection (SFFS) (wrapper type) with the k nearest neighbors (kNN) algorithm. Selected features for Gusty wind speed at every band are the maximum and mean wind speed at the 2 and 10 minutes, and the momentary wind speed by the MI method. And selected by SFFS method is the wind direction deviation in past 10 minutes at bands number 11 and Mid, momentary pressure at Mid band and maximum wind speed at 10 minutes in band 29. But for Gusty wind direction by first method: the selected features are minimum, mean and maximum wind direction at 2 minutes, minimum and mean wind direction at 10 minutes and momentary wind direction in band 29. And with second method, they are the wind direction deviations at past 10 minutes at the bands 29 and Mid, and the mean sea level pressure and mean wind direction at 10 minutes in band 29.
In the final step, these selected features were used as inputs to the multilayer perceptron neural network in different modes such as: layer number, neuron number, learning rate and threshold value for weight of neuron. And the model output results were compared to predict the gusty direction and speed. Then the best model selected. The results show that to predict the wind speed, the Best model is a multilayer perceptron neural network with two-layer , 4 neurons in the first layer, 2 neurons in the second layer and 1 neuron in the output layer, learning rate equals 0.1 and initial weight Neurons equal 0.5. But for predict the wind direction; the Best model is two-layer, 6 neurons in the first layer and 3 neurons in the second layer with the same learning rate and initial threshold. The MLP performance is better in predicting Gusty wind speed.